\n\n\n \n \n\n\n# exceptd Security\n\n**AI security skills grounded in mid-2026 threat reality, not framework documentation from 2020.**\n\n[](https://github.com/blamejs/exceptd-skills/releases)\n[](https://www.npmjs.com/package/@blamejs/exceptd-skills)\n[](https://github.com/blamejs/exceptd-skills/actions/workflows/ci.yml)\n[](https://scorecard.dev/viewer/?uri=github.com/blamejs/exceptd-skills)\n[](https://www.apache.org/licenses/LICENSE-2.0)\n[](#skill-inventory)\n[](https://atlas.mitre.org)\n[](https://attack.mitre.org)\n[](AGENTS.md)\n[](data/global-frameworks.json)\n\n
\n\n---\n\n**Core premise:** Every major security and compliance tool on the market is still operating on stale threat models. NIST 800-53, ISO 27001, SOC 2, and PCI-DSS were written for network-centric, on-prem or early-cloud environments. They have no controls for AI pipeline integrity, MCP/agent tool trust boundaries, LLM prompt injection as an access control failure, page-cache exploitation bypassing filesystem integrity checks, or ephemeral infrastructure where traditional asset inventory is architecturally impossible.\n\nEvery skill explicitly flags where a compliance framework's control is insufficient for current attack patterns. The framework is often the problem, not the org.\n\n## What ships\n\nPre-1.0. Released on [GitHub](https://github.com/blamejs/exceptd-skills/releases)\nand on npm as [`@blamejs/exceptd-skills`](https://www.npmjs.com/package/@blamejs/exceptd-skills),\nwith npm provenance attestation and an Ed25519 signature on every skill body.\n\n51 skills across kernel LPE, MCP supply chain, AI-as-C2, prompt injection,\npost-quantum crypto, SBOM integrity, identity incident response and 35 further\ndomains.\n\n| | |\n|---|---|\n| **11 catalogs** | CVE, ATLAS, ATT&CK, CWE, D3FEND, DLP, RFC, framework gaps, global frameworks, zero-day lessons, exploit availability — across 35 jurisdictions |\n| **1,323 CVEs** | Each with behavioral indicators, an ATT&CK technique mapping, and a defense-chain zero-day lesson |\n| **33 playbooks** | Seven-phase investigations: `kernel`, `mcp`, `ai-api`, `sbom`, `containers`, `crypto`, `secrets` and the rest, through `webhook-callback-abuse`, `cicd-pipeline-compromise`, `identity-sso-compromise`, `llm-tool-use-exfil` and `post-quantum-migration`. `exceptd brief --all` is the live list |\n| **A CLI** | `discover → brief → run → attest`, where each run executes the playbook's seven-phase contract |\n\nA nightly job pulls KEV, NVD, EPSS, GHSA, OSV and IETF deltas alongside 15\nadvisory, vendor-research and tech-press feeds, and opens a pull request for\neditorial review rather than merging. A companion watcher flags historical CVEs\nre-broken without a new identifier being assigned.\n\n---\n\n## Skill Inventory\n\nSeventeen of the fifty-one shipped skills are described below. Run `exceptd skill`\nwith no argument for the full list with one-line summaries, or `exceptd skill `\nfor one skill's description, triggers and data dependencies; the body itself is\n`skills//skill.md`. Skills named elsewhere in this file but absent here —\n`pqc-first` and `security-maturity-tiers` among them — ship too; this section is a\nselection, not the inventory.\n\n### Triage & Dispatch\n\n**[researcher](skills/researcher/skill.md)**\nFront-door triage skill for raw threat intel. Takes a CVE ID, ATLAS TTP, vendor advisory, framework control ID, or incident narrative; cross-joins it across `data/cve-catalog.json`, `data/atlas-ttps.json`, `data/framework-control-gaps.json`, `data/zeroday-lessons.json`, `data/exploit-availability.json`, and `data/global-frameworks.json`; produces a one-page RWEP-anchored dispatch report; routes the operator to the right specialized skill(s). Start here when the input is \"here's a thing, tell me what to do with it\".\n\n### Kernel & Privilege Escalation\n\n**[kernel-lpe-triage](skills/kernel-lpe-triage/skill.md)**\nAssess Linux kernel local privilege escalation exposure. Covers Copy Fail (CVE-2026-31431, CISA KEV, 732-byte deterministic root, all Linux since 2017), Dirty Frag (CVE-2026-43284/CVE-2026-43500, page-cache chain via ESP/IPsec and RxRPC). Outputs: exposure score, live-patch vs. reboot remediation path, compensating controls, framework gap declaration.\n\n### AI-Specific Attack Surface\n\n**[ai-attack-surface](skills/ai-attack-surface/skill.md)**\nAI/ML attack surface assessment mapped to MITRE ATLAS v2026.09 with explicit gap flags. Covers prompt injection as enterprise RCE (CVE-2025-53773 CVSS 7.8, 85%+ bypass rate against SOTA defenses), MCP supply chain RCE (CVE-2026-30615, zero user interaction, 150M+ downloads), RAG exfiltration, model poisoning, AI-assisted exploit development (41% of 2025 zero-days), credential theft acceleration (160% increase).\n\n**[mcp-agent-trust](skills/mcp-agent-trust/skill.md)**\nEnumerate MCP (Model Context Protocol) trust boundary failures. Covers tool allowlisting gaps, unsigned server manifests, prompt injection via tool responses, supply chain compromise. CVE-2026-30615 (Windsurf, zero-interaction RCE). Generates: tool allowlist policy, server signing requirements, bearer auth config, output sanitization requirements.\n\n**[rag-pipeline-security](skills/rag-pipeline-security/skill.md)**\nRAG-specific threat model with no current framework coverage. Embedding manipulation for data exfiltration, vector store poisoning, chunking attacks, retrieval filter bypass, indirect prompt injection via retrieved documents. ATLAS-mapped. Generates: retrieval audit controls, anomaly detection requirements, output monitoring policy.\n\n**[ai-c2-detection](skills/ai-c2-detection/skill.md)**\nDetect adversary use of AI APIs as covert command-and-control (SesameOp case study, ATLAS AML.T0096). PROMPTFLUX/PROMPTSTEAL malware families that query LLMs during execution for real-time evasion. Outputs: behavioral baseline model, detection signatures, network monitoring rules, incident response playbook.\n\n### Framework & Compliance\n\n**[framework-gap-analysis](skills/framework-gap-analysis/skill.md)**\nFeed a compliance framework control ID and a threat scenario — receive: what the control was designed for, why it is insufficient against current TTPs, which attacker technique exploits the gap, what a real control would require. Built-in gap mappings for NIST 800-53, ISO 27001:2022, SOC 2, PCI-DSS 4.0, NIS2, DORA, CIS v8.\n\n**[compliance-theater](skills/compliance-theater/skill.md)**\nIdentify where an organization passes an audit but remains exposed. Seven documented compliance theater patterns with specific detection tests. Outputs: theater score per control domain, exposure summary, auditor-facing remediation language, evidence gap list.\n\n**[global-grc](skills/global-grc/skill.md)**\nMulti-jurisdiction GRC mapping. Covers EU (GDPR Art. 32, NIS2, DORA, EU AI Act, EU CRA), UK (Cyber Essentials Plus, NCSC CAF), Australia (ISM, ASD Essential 8, APRA CPS 234), Singapore (MAS TRM, CSA CCoP), Japan (METI, NISC), India (CERT-In, SEBI), Canada (OSFI B-10), and global (ISO 27001:2022, CSA CCM v4, CIS Controls v8). Identifies universal gaps that no jurisdiction's framework covers.\n\n**[policy-exception-gen](skills/policy-exception-gen/skill.md)**\nGenerate defensible policy exceptions for architectural realities frameworks don't accommodate. Templates for: ephemeral/serverless infrastructure (no traditional asset inventory), AI pipelines (continuous opaque model updates), zero trust architecture (no network perimeter), live-system no-reboot patching. Each exception includes compensating controls, risk acceptance language, and auditor-ready justification.\n\n### Risk Intelligence\n\n**[exploit-scoring](skills/exploit-scoring/skill.md)**\nReal-World Exploit Priority (RWEP) scoring beyond CVSS. Factors: CISA KEV status (0.25), public PoC (0.20), AI-assisted weaponization (0.15), active exploitation (0.20), patch availability (-0.15), live-patch availability (-0.10), blast radius (0.30). Pre-calculated RWEP scores for all CVEs in `data/cve-catalog.json`. Outputs RWEP alongside CVSS with plain-language priority guidance.\n\n**[threat-model-currency](skills/threat-model-currency/skill.md)**\nScore how current an organization's threat model is against 2026 threat reality. Checklist of 14 current threat classes against documented model coverage. Outputs: currency percentage, specific missing threat classes, recommended additions with ATLAS/ATT&CK references, prioritized update roadmap.\n\n**[zeroday-gap-learn](skills/zeroday-gap-learn/skill.md)**\nRun the zero-day learning loop: zero-day description → attack vector extraction → control gap identification → framework coverage assessment → new control requirement generation → exposure scoring. Encodes lessons from Copy Fail, Dirty Frag, CVE-2025-53773, CVE-2026-30615, SesameOp. Feeds back into framework-gap-analysis and threat-model-currency.\n\n### Identity, OT, Disclosure & Threat Modeling\n\n**[identity-assurance](skills/identity-assurance/skill.md)**\nIdentity assurance for mid-2026. NIST 800-63 AAL/IAL/FAL levels, FIDO2/WebAuthn passkey deployment, OIDC/SAML/SCIM federation, agent-as-principal identity for autonomous AI workloads, short-lived workload token issuance, OAuth 2.0 + RFC 9700 (OAuth 2.0 Security BCP) hardening. Outputs: assurance-level gap map, passkey rollout plan, agent identity policy, token-lifetime targets.\n\n**[ot-ics-security](skills/ot-ics-security/skill.md)**\nOT / ICS security for mid-2026. NIST 800-82r3, IEC 62443-3-3, NERC CIP, IT/OT convergence risks (flat networks, shared AD, jump-host weaknesses), AI-augmented HMI threats, and ATT&CK for ICS mappings. Outputs: zone/conduit gap map, safety-instrumented-system isolation review, OT-specific patching exception templates.\n\n**[coordinated-vuln-disclosure](skills/coordinated-vuln-disclosure/skill.md)**\nCoordinated Vulnerability Disclosure for mid-2026. ISO 29147 (disclosure) + ISO 30111 (handling), VDP and bug bounty design, CSAF 2.0 machine-readable advisories, security.txt (RFC 9116), EU CRA / NIS2 regulator-mandated disclosure timelines, AI-specific vulnerability classes (prompt injection, training data poisoning, model exfiltration). Outputs: VDP policy, advisory template, regulator notification calendar.\n\n**[threat-modeling-methodology](skills/threat-modeling-methodology/skill.md)**\nMethodology selection and execution across STRIDE, PASTA, LINDDUN (privacy), Cyber Kill Chain, Diamond Model, MITRE Unified Kill Chain, AI-system threat modeling, and agent-based threat modeling. Outputs: methodology choice with justification, scoped DFD or attack tree, threat-to-control crosswalk against ATLAS / ATT&CK / D3FEND.\n\n---\n\n## Install\n\nThree audience paths. Pick the one that matches how you'll use this.\n\n### 1. AI consumer (read-only — most users)\n\nYou want an AI assistant to load the skills + catalogs against a question of yours. Easiest path:\n\n```bash\nnpx @blamejs/exceptd-skills path\n```\n\nThat prints the absolute path of the installed package. Point your AI assistant at:\n\n- `/AGENTS.md` — canonical project rules + ground truth for every skill\n- `/data/_indexes/summary-cards.json` — 100-word abstract per skill (~95 KB)\n- `/data/_indexes/recipes.json` — curated multi-skill chains for common use cases\n\nNo clone, no signing keys, no Node 24 required for assistants that read directly from disk. If your assistant needs a local copy as a regular checkout, use `npx degit blamejs/exceptd-skills my-skills` instead.\n\n### 2. Operator (run commands locally)\n\nYou want to refresh CVE/RFC data, run currency checks, or generate reports. Install + invoke via `npx` (no global install needed):\n\n```bash\nnpx @blamejs/exceptd-skills doctor # health check\nnpx @blamejs/exceptd-skills refresh --apply --swarm # pull KEV/NVD/EPSS/RFC/GHSA + apply\nnpx @blamejs/exceptd-skills refresh --advisory CVE-2026-45321 # seed one CVE draft from GHSA\nnpx @blamejs/exceptd-skills refresh --advisory MAL-2026-3083 # seed via OSV (MAL-/SNYK-/RUSTSEC-/USN-/PYSEC-/GO-/MGASA-/UVI-)\nnpx @blamejs/exceptd-skills refresh --curate CVE-2026-45321 # surface editorial questions for a draft\nnpx @blamejs/exceptd-skills refresh --network # swap data/ from latest signed npm tarball\n```\n\nFor frequent use, install globally to skip the `npx` resolution every time:\n\n```bash\nnpm install -g @blamejs/exceptd-skills\nexceptd help\n```\n\nFirst run — verify the signing chain and pin the public-key fingerprint for out-of-band checks:\n\n```bash\nexceptd doctor --signatures # verify Ed25519 chains (51/51 expected)\ncat $(exceptd path)/keys/EXPECTED_FINGERPRINT # pin fingerprint for OOB verify\n```\n\nVerify on npm: `npm view @blamejs/exceptd-skills@ dist.signatures` shows the SLSA v1 provenance attestation.\n\nAir-gapped operation: run `exceptd refresh --prefetch` on a connected host, copy the resulting `.cache/upstream/` to the airgap, run `exceptd refresh --from-cache --apply` over there. The vendored upstream snapshots replace every network call.\n\nOptional env vars for higher rate budgets:\n\n| Variable | Purpose |\n|---|---|\n| `NVD_API_KEY` | Lifts NVD 2.0 from 5 → 50 requests per 30s window. Free key at . |\n| `GITHUB_TOKEN` | Lifts GitHub Releases + GHSA from 60 → 5000 requests per hour. |\n| `EXCEPTD_GHSA_FIXTURE` | Path to a JSON fixture matching the api.github.com/advisories shape. For offline tests + air-gap workflows. |\n| `EXCEPTD_OSV_FIXTURE` | Path to a JSON fixture matching the OSV schema (https://ossf.github.io/osv-schema/). For offline tests + air-gap workflows against the OSV source (added v0.12.10). |\n| `EXCEPTD_REGISTRY_FIXTURE` | Path to a JSON fixture matching the npm registry response. Used by `doctor --registry-check` + `run --upstream-check` + `refresh --network` for offline testing. |\n\n### 3. Maintainer (extend / sign / publish)\n\nYou're adding a skill, updating a catalog, or cutting a release. Clone + bootstrap the full toolchain:\n\n```bash\ngit clone https://github.com/blamejs/exceptd-skills\ncd exceptd-skills\nnpm run bootstrap # auto-detects: verify-only / re-sign / first-init\nnpm run predeploy # full predeploy gate sequence locally\n```\n\n`bootstrap` auto-detects the right mode based on which keys exist on disk:\n\n- **Verify-only** (default on a fresh clone): `keys/public.pem` ships in the repo, no `.keys/private.pem` locally. Checks that every skill verifies against the shipped signature, exits.\n- **Re-sign**: `.keys/private.pem` exists locally. Re-signs every skill against current content, verifies.\n- **First-init**: no `keys/public.pem` shipped or `--init` passed. Generates a new Ed25519 keypair, signs everything.\n\nDirect invocations also available: `npm run verify`, `node lib/sign.js sign-all`.\n\n## Keeping the catalog current\n\nA CVE reaches the catalog through one of three paths, at different speeds.\n\n**The nightly job** pulls KEV, NVD, EPSS, IETF, GHSA and OSV. They arrive at\ndifferent times: KEV takes days and NVD around ten, while GHSA fires within hours\nof disclosure across npm, PyPI, Maven, Go and NuGet. OSV layers Snyk, RustSec,\nMageia, Ubuntu USN, the Go vulnerability database, PYSEC and UVI on top, plus the\nOSSF Malicious Packages dataset — which is where a compromised package with no\nCVE shows up at all. New IDs land as drafts flagged `_auto_imported`, which the\ncatalog validator reports as warnings rather than errors: the entry is available\nimmediately and the editorial work follows.\n\n```bash\nexceptd refresh --advisory CVE-2026-45321 --apply # pull one ID now\nexceptd refresh --curate CVE-2026-45321 # editorial questions for a draft\nexceptd refresh --drift-only --apply # reconcile what is already held\n```\n\n**Primary-source polling** runs ahead of all of them. `exceptd refresh\n--check-advisories` reads 15 feeds — eight advisory and coordinated-disclosure\nvenues (Qualys TRU, Red Hat RHSA, Ubuntu USN, ZDI, kernel.org commits,\noss-security, JFrog SecOps, CISA), four vendor research blogs (Microsoft\nSecurity, Sysdig, Trail of Bits, Embrace the Red), and three press and researcher\ntrackers. Between them they publish CVE IDs three to fourteen days ahead of NVD\nenrichment. The command is report-only: it returns each newly-seen ID with its\nsources and advisory URLs and changes nothing. `lib/cve-regression-watcher.js`\nreads the same diffs for references to CVEs two or more years old, which is what\na silently re-broken historical flaw looks like before anyone assigns it a new\nidentifier.\n\n**A daily agent** (`routine: exceptd-threat-intake`, 14:00 UTC) chains those\nsteps — poll, alert, refresh, enrich up to five new IDs, re-sign and rebuild if\nthe catalog moved — and commits to an `intake/` branch. Managed at\n.\n\nOnce a refresh lands, `exceptd watchlist --alerts` matches the catalog against\nfive operational patterns — `kernel_lpe_with_poc`, `supply_chain_family`,\n`ai_discovered_kev`, `active_exploitation_unpatched`, `recent_poc_no_kev_yet` —\nsorted critical-first, then by RWEP, so a refreshed catalog can be triaged\nwithout reading every entry.\n\n## Evidence collection\n\n`exceptd collect ` walks the working directory, applies the playbook's\ncatalogued patterns, stats permissions, and emits submission JSON in the shape\n`exceptd run --evidence -` accepts:\n\n```bash\nexceptd collect | exceptd run --evidence -\nexceptd doctor --collectors # which playbooks ship one, and which do not\n```\n\n14 of the 33 playbooks ship a collector. The other 19 are deliberately without\none: they are judgement-shaped incident, governance and pure-analyze playbooks\nwhere the evidence is an assessment rather than a file on disk.\n`cicd-pipeline-compromise` additionally requires `--attest-ownership`, because\nits `operator-owns-ci-fleet` precondition is what stops it being pointed at\nsomeone else's CI.\n\nTwo scans sit outside the playbook contract:\n\n- `exceptd watchlist --org-scan --org ` searches one GitHub org for\n repositories matching known threat-actor naming patterns. This is the\n detection for the supply-chain class that uses GitHub itself as the exfil\n channel. Set `GITHUB_TOKEN` for private repositories and rate-limit headroom.\n- `exceptd doctor --ai-config` checks `~/.claude`, `~/.cursor`, `~/.codeium`,\n `~/.aider` and `~/.continue` for credential-bearing files — `settings.json`,\n `mcp.json`, `api_key*`, `*.token`, `*.credentials` — left readable beyond mode\n 0600. Windows has no equivalent bit, so findings there are raised for manual\n ACL review. Opt-in; the bare `doctor` pass does not run it.\n\n## CLI command reference\n\nEvery command works the same via `npx @blamejs/exceptd-skills`, a global install (`exceptd`), or a local `node bin/exceptd.js`.\n\n### Canonical verbs\n\n```\nexceptd First-run welcome — two ways to start\n (discover / ask) plus common starting\n playbooks for code / Linux / service contexts.\n\nexceptd discover Scan cwd → recommend playbooks based on\n detected files (.git, package.json,\n Dockerfile, requirements.txt, etc) + host\n platform. Replaces scan + dispatch.\n --scan-only Also include legacy host scan findings.\n --json | --pretty Machine output (default is human checklist).\n\nexceptd brief [playbook] Unified info doc — jurisdictions + threat\n context + RWEP thresholds + preconditions\n + artifacts + indicators. Replaces plan +\n govern + direct + look.\n --all Every playbook (replaces `plan`).\n --scope system | code | service | cross-cutting.\n --directives Expand directive metadata per playbook.\n --phase Emit only one phase (legacy compat).\n\nexceptd run [playbook] Phases 4-7. Auto-detects cwd context when\n no playbook positional.\n --evidence Submission JSON (flat or nested shape).\n --evidence-dir Per-playbook submission files (cron-friendly).\n --scope | --all Multi-playbook run.\n --vex CycloneDX / OpenVEX filter (drop not_affected).\n --format ... csaf-2.0 | sarif | openvex | markdown | summary | json.\n Repeatable. CSAF is primary; extras go to\n close.evidence_package.bundles_by_format.\n --diff-from-latest Drift vs prior attestation for same playbook.\n --ci Exit-code gate (use `exceptd ci` instead).\n --operator Bind attestation to identity.\n --ack Explicit jurisdiction-obligation consent.\n --session-id Reuse session id (collision refused).\n --force-overwrite Override session collision refusal.\n --session-key HMAC sign evidence_package (≥ 16 hex chars).\n --attestation-root Override ~/.exceptd/attestations/ root.\n Alternative: set EXCEPTD_HOME=\n env var (attestations land in\n $EXCEPTD_HOME/attestations/). Useful for\n multi-tenant shared hosts where each\n operator wants a private attestation\n root, or for CI runners that should\n scope attestations to the job workspace.\n --explain Dry-run: preconditions + artifacts +\n signal keys + submission skeleton.\n --signal-list Lighter than --explain; enumerate signal\n keys only.\n --force-stale Override threat_currency_score < 50 gate.\n --air-gap Honor air_gap_alternative paths.\n\nexceptd ai-run JSONL streaming variant of run. AI emits\n evidence events on stdin; runner streams\n phase events on stdout. One pipe, no\n file handoff. See `exceptd ai-run --help`\n for the full stdin event grammar.\n --no-stream Single-shot mode (emit one combined JSON).\n\n# Stdin event the host emits (one JSON object per line):\n# {\"event\":\"evidence\",\"payload\":{\n# \"precondition_checks\": {...}, // per-precondition boolean assertions\n# \"observations\": {...}, // per-artifact + per-indicator captures\n# \"verdict\": {...} // optional operator-supplied verdict\n# }}\n# observations[] carries both artifact captures\n# ({ captured: true, value: \"...\" }) AND indicator overrides\n# ({ indicator: \"\", result: \"hit\"|\"miss\" }) — the runner normalises\n# both branches from a single map. The alternative nested shape\n# ({ artifacts, signal_overrides, signals }) is also accepted, but do not\n# mix the two — if `signal_overrides` is present, `observations` and\n# `verdict` are ignored.\n# Phases emitted on stdout (in order): govern → direct → look →\n# await_evidence → detect → analyze → validate → close → done.\n# Errors emit {\"event\":\"error\",\"reason\":\"...\"} and exit non-zero.\n\nexceptd collect Walk cwd + invoke the companion collector\n under lib/collectors/.js. Emits\n a submission JSON ready to pipe into\n `exceptd run --evidence -`.\n 14/33 playbooks have collectors; the rest\n are AI-driven by design (incident /\n governance / pure-analyze — see\n AGENTS.md).\n --cwd Collect against a different repo / host.\n --pretty Indented JSON.\n --attest-ownership cicd-pipeline-compromise only — opt-in to\n the operator-owns-ci-fleet precondition\n so the runner doesn't halt at preflight.\n\n# Canonical operator flow on a freshly-cloned repo:\nexceptd discover # which playbooks apply here?\nexceptd collect | exceptd run --evidence - # full pipe to verdict\nexceptd doctor --collectors # list every collector + which are skipped\n\nexceptd attest [] Auditor-facing operations.\n attest list Inventory all sessions across both\n ~/.exceptd and cwd-legacy roots.\n attest show Full (unredacted) attestation.\n attest export Redacted bundle for audit submission.\n Strips raw artifact values; preserves\n evidence_hash + signature + verdict.\n --format csaf wraps in CSAF envelope.\n attest verify Ed25519 .sig sidecar verification.\n attest diff Drift replay (= reattest default).\n --against compares two\n sessions side-by-side with per-artifact\n diff (added / removed / changed).\n --playbook Filter (list / diff).\n --since Filter list / diff to entries after date.\n\nexceptd discover / doctor / ci See above for doctor and ci.\n\nexceptd doctor One-shot health check.\n --signatures Only Ed25519 skill verification.\n --currency Only skill currency report.\n --cves Only CVE catalog drift check.\n --rfcs Only RFC catalog drift check.\n --ai-config Audit AI-assistant config-file permissions\n across ~/.claude, ~/.cursor, ~/.codeium,\n ~/.aider, ~/.continue. Flags sensitive\n files (settings.json, mcp.json,\n *.mcp_config.json, api_key*, *.token,\n *.credentials) not at mode 0600 on POSIX;\n surfaces an info-level \"manual ACL review\"\n note for each sensitive file on Windows.\n Opt-in; not part of the default doctor\n pass.\n --fix Auto-remediate signing gaps: regenerate\n the local Ed25519 private key when\n keys/public.pem exists but .keys/private.pem\n is absent. No-op when the key is present.\n --registry-check Probe the npm registry for the latest\n published version + days-since-publish.\n Off by default; --air-gap suppresses it.\n --collectors Enumerate the per-playbook collector layer:\n which playbooks ship a collector, which are\n policy-skipped, and which are unwired.\n --shipped-tarball Run the pack + extract + verify round-trip\n against the tarball operators receive, not\n just the source tree.\n --exit-codes Print the canonical exit-code table as\n JSON for CI / scripting consumers.\n\nexceptd ci One-shot CI gate. Exit codes: 0 PASS,\n 1 framework error, 2 detected/escalate\n (or rwep ≥ rwep_threshold.escalate),\n 3 ran-but-no-evidence, 4 blocked\n (ok:false), 5 jurisdiction clock started.\n --all | --scope Pick playbooks; auto-detect if neither.\n --max-rwep Cap below playbook default.\n --block-on-jurisdiction-clock Fail when notification clock fires.\n --evidence / --evidence-dir Per-playbook submission files.\n\nexceptd ask \"\" Plain-English routing to playbook(s).\n Returns ranked playbook IDs based on\n keyword overlap with each playbook's\n domain.name + attack_class + threat_context.\n A question in a domain covered by a skill\n rather than a playbook (email-auth, child\n safety, HIPAA, DLP) surfaces the skill.\n\nexceptd recipes [] List the curated multi-skill workflows;\n `recipes ` expands one into its\n ordered skill chain.\n\nexceptd lint Pre-flight check submission shape vs\n playbook (preconditions / artifacts /\n indicators) without executing phases 4-7.\n\nexceptd cve Resolve one CVE citation → status\n (published / rejected / disputed /\n fabricated / nonexistent / unknown) plus\n cvss / kev / product. Order: curated\n catalog (offline) → resolved cache\n (7-day TTL, warmed by a prior lookup) →\n one NVD lookup, then cached. Lets a\n fan-out of agents share one answer\n instead of each researching the same id.\n --air-gap | --no-network Offline-only (also EXCEPTD_AIR_GAP=1).\n Returns unknown + a reason when the id\n isn't in catalog/cache.\n --json | --pretty Machine output.\n Exit 2 when the citation won't stand up\n (rejected / fabricated / nonexistent /\n withdrawn).\n\nexceptd rfc Resolve an RFC number → title + status\n from the local index (whole current\n series, fully offline).\n --check \"\" Report title_match true/false; exit 2 on\n mismatch (e.g. RFC 9404 cited as the\n Sieve spec — it's JMAP Blob Management).\n --air-gap Offline-only. Not-found numbers are\n likely obsoleted/historic or nonexistent;\n with network it disambiguates via the\n datatracker.\n --json | --pretty Machine output.\n\nexceptd refresh Refresh upstream catalogs + indexes.\n Replaces prefetch + refresh + build-indexes.\n --apply Write diffs back + rebuild indexes.\n --from-cache [<dir>] Read from prefetch cache.\n --prefetch Warm the offline cache by fetching every\n upstream artifact now (network required).\n Run on a connected host, then point\n --from-cache at the result on the air-gap.\n --drift-only Reconcile the entries the catalog already\n holds; skip discovery of new ones. Use it\n to correct stale fields on shipped entries\n without pulling in drafts that still need\n curation.\n --no-network Report-only dry-run: list what would be\n fetched without touching the network.\n --network (v0.11.14) Fetch latest signed catalog\n snapshot from npm tarball, verify against\n local public.pem, swap data/ in place.\n --advisory <CVE-or-GHSA-ID> (v0.12.0) Seed a single catalog entry from\n GitHub Advisory Database. Writes a draft\n flagged _auto_imported. --apply commits it.\n --curate <CVE-ID> (v0.12.0) Emit editorial questions + ranked\n candidates (ATLAS/ATT&CK/CWE/framework) for\n a draft catalog entry.\n --check-advisories Poll 15 primary-source advisory feeds\n (Qualys TRU, Red Hat RHSA, Ubuntu USN,\n ZDI, kernel.org commits, oss-security\n mailing list, JFrog SecOps, CISA current\n advisories, Microsoft Security Blog,\n Sysdig, Trail of Bits, Embrace the Red,\n BleepingComputer, The Hacker News,\n researcher activity-feed tracker) for\n CVE IDs disclosed at T+0 to T+1 —\n days ahead of NVD enrichment.\n Report-only: emits structured diffs[]\n with {cve_id, sources[], advisory_urls[],\n disclosed_at, title}; does NOT mutate the\n catalog. Route promising IDs through\n `refresh --advisory <CVE-ID>` to enrich.\n --indexes-only Rebuild data/_indexes/*.json only.\n\nSources (default = all): kev | epss | nvd | rfc | pins | ghsa | osv.\nGHSA covers npm, PyPI, Maven, Go, NuGet, etc.; OSV layers Snyk, RustSec,\nMageia, Ubuntu USN, Go Vuln DB, PYSEC, UVI, plus the OSSF Malicious\nPackages dataset (`MAL-*` keys). New IDs land as drafts that the catalog\nvalidator treats as warnings, not errors — editorial review (framework\ngaps, IoCs, ATLAS/ATT&CK refs) is still required.\n\nexceptd watchlist Default mode: aggregate every skill's\n forward_watch entries (upcoming standards,\n RFC publications, new TTPs to monitor) in\n one shot.\n `--by-skill` inverts the grouping.\n --alerts Switch to CVE-catalog pattern alerts.\n Five patterns ship:\n - kernel_lpe_with_poc (high) — kernel\n LPE class with public PoC + blast\n radius >= 25\n - supply_chain_family (high) — MAL-*\n entries or `type: malicious-*`\n - ai_discovered_kev (high) — AI-\n discovered AND CISA KEV-listed\n - active_exploitation_unpatched\n (critical) — confirmed in-the-wild\n + no patch available\n - recent_poc_no_kev_yet (medium) —\n public PoC verified within 14 days,\n not yet KEV-listed\n Sorted critical-severity first, then by\n RWEP descending. JSON or human output.\n --org-scan --org <login> Probe GitHub Search for repositories\n matching known threat-actor naming\n patterns (\"A Gift From TeamPCP\",\n \"Shai-Hulud\", \"TeamPCP\") scoped to one\n org. Custom patterns via repeatable\n `--pattern <s>`. Set GITHUB_TOKEN for\n private-repo coverage + higher rate\n limit; without it, public-repo search\n only.\n\nexceptd watch Long-running forward-watch daemon. Blocks\n and listens for KEV additions, ATLAS\n updates, CVE drops, and framework\n amendments, with scheduled currency /\n validation checks. Ctrl-C (or SIGTERM /\n SIGHUP / SIGBREAK) to stop. For one-shot\n aggregation, pattern alerts, or org-scan,\n use `exceptd watchlist`.\n\nexceptd skill <name> Show context for one skill.\nexceptd framework-gap <FW> <ref> One framework + one CVE/scenario, JSON\n or human. (Operates outside the seven-\n phase contract for ad-hoc gap analysis.)\nexceptd report [executive] Structured posture report. Bare `report`\n emits the full posture; the optional\n `executive` argument emits the\n executive-summary view.\nexceptd path Absolute path to the installed package.\nexceptd version Package version.\nexceptd help This help.\nexceptd <verb> --help Most verbs print per-verb usage with flag\n descriptions.\n```\n\n### Deprecated and removed verbs\n\nFive verbs removed in v0.13.0 after deprecation since v0.11.0. Invoking any of these now returns a structured `ok:false` refusal pointing at the replacement; pre-v0.13 scripts must migrate.\n\n| Removed verb | Replacement |\n|---|---|\n| `plan` | `brief --all` |\n| `govern <pb>` | `brief <pb> --phase govern` |\n| `direct <pb>` | `brief <pb> --phase direct` |\n| `look <pb>` | `brief <pb> --phase look` |\n| `ingest` | `run` |\n\nThe remaining v0.10.x verbs are still functional, no banner, no removal scheduled. Two shapes:\n\n**Canonical-equivalent aliases** — same output shape as the canonical verb; safe to use interchangeably:\n\n| Alias | Canonical | Output shape |\n|---|---|---|\n| `verify` | `doctor --signatures` | matches canonical |\n| `validate-cves` | `doctor --cves` | matches canonical |\n| `validate-rfcs` | `doctor --rfcs` | matches canonical |\n| `list-attestations` | `attest list` | matches canonical |\n| `reattest <sid>` | `attest diff <sid>` | matches canonical |\n| `prefetch` | `refresh --prefetch` | matches canonical |\n| `build-indexes` | `refresh --indexes-only` | matches canonical |\n\n**Legacy passthrough verbs** — dispatch to the v0.10.x orchestrator script. The output shape is **NOT** identical to the canonical verb — it's the legacy `{timestamp, host, findings}` envelope. Use the canonical verb when you want the v0.11+ structured envelope contract; the passthrough is kept only for scripts that depend on the legacy output:\n\n| Passthrough | Canonical (different output shape) |\n|---|---|\n| `scan` | `discover --scan-only` |\n| `dispatch` | `discover` |\n| `currency` | `doctor --currency` |\n\n### Result envelope contract\n\nEvery `run` (and every per-playbook result inside a `ci` body) hoists the headline summary fields to the top of the JSON envelope so machine consumers do not have to walk `phases.*` to find them:\n\n| Field | Type | Meaning |\n|---|---|---|\n| `ok` | boolean | `true` on success, `false` on blocked-at-preflight or persistence failure |\n| `playbook_id` | string | Playbook id (present on blocked results too, so a `results[]` iterator can identify the row without joining against `playbooks_run[]` by index) |\n| `directive_id` | string | Directive within the playbook |\n| `session_id` | string | Run id (used by `attest verify <sid>` / `attest diff <sid>`) |\n| `verdict` | string | One of `detected` / `not_detected` / `inconclusive` / `pending` / `skipped` / `blocked` |\n| `rwep_score` | number \\| null | `phases.analyze.rwep.adjusted`, or `null` on blocked / catalog-baseline-zero runs |\n| `top_finding` | string \\| null | First matched CVE id, or the indicator classification when no CVE correlated |\n| `summary_line` | string | One-line human summary (~240 chars) — `<playbook>: <verdict> (rwep=<n>, <finding>, evidence=<state>)` |\n| `evidence_completeness` | string | One of `complete` / `partial` / `missing` / `unknown` / `not-evaluated` |\n| `indicators_evaluated` | number \\| null | Indicators that produced a verdict |\n| `indicators_known` | number \\| null | Indicators declared by the playbook |\n| `evidence_hash` | string | SHA-256 of the normalized submission |\n| `submission_digest` | string | SHA-256 of the structured envelope |\n| `attestation_path` | string | Absolute path to the persisted attestation JSON (success path only) |\n| `preflight_issues` | array | Preconditions evaluated, with per-precondition `on_fail` + `check` |\n| `precondition_check_source` | object | Per-precondition: `submission` / `runOpts` / `merged` |\n| `phases` | object | Full per-phase outputs — `govern`, `direct`, `look`, `detect`, `analyze`, `validate`, `close` |\n\nOn a blocked result (preflight halt, missing precondition), `ok` is `false` and the envelope additionally carries `blocked_by` / `reason` / `remediation` / `phase: 'preflight'` / `verdict: 'blocked'`. `evidence_completeness` reports `not-evaluated`. In default human output a blocked result renders as a one-line `[blocked]` summary with the reason and a next step; `--json` / `--pretty` return the full envelope.\n\n### Default terminal output vs `--json` / `--pretty`\n\nBy default `ci`, `run`, `attest verify`, `attest diff`, and `discover` emit a human-readable digest at the terminal — verdict line, per-playbook table (for `ci`), next-step block keyed on verdict (BLOCKED → `exceptd lint <pb> -`; NO_EVIDENCE → lint + `--evidence-dir`; FAIL → `--format markdown` / `--format csaf-2.0` per detected playbook; CLOCK_STARTED → CSAF advisory), pending jurisdiction obligations grouped by `clock_start_event`, deduped session warnings, framework-gap rollup.\n\nPass `--json` (compact) or `--pretty` (indented) to reach the structured envelope when automating. Setting `EXCEPTD_RAW_JSON=1` in the environment has the same effect. `--quiet` keeps human output but drops advisory stderr notes (and the deprecation / unsigned-attestation banners) so `run … 2>&1 | jq` stays clean; `--json-stdout-only` goes further and silences all stderr.\n\n## Invoking a skill from your AI assistant\n\nOnce your assistant has loaded `AGENTS.md`, type a trigger phrase or skill name:\n\n```\nkernel-lpe-triage\nai-attack-surface\nframework-gap-analysis NIST-800-53-SI-2 CVE-2026-31431\ncompliance-theater\nglobal-grc NIS2\nexploit-scoring CVE-2026-31431\nzeroday-gap-learn CVE-2026-30615\nsecurity-maturity-tiers\npqc-first\n```\n\n## AI assistant configuration\n\nThe canonical agent-agnostic project rules live in `AGENTS.md` — the **only** project-rules file in this repo. The project does not ship per-vendor mirrors; each tool is configured to load `AGENTS.md` directly.\n\n| Assistant | How it picks up the rules |\n|-----------|---------------------------|\n| OpenAI Codex CLI, Sourcegraph amp, Aider, Continue, Cline, Roo Code, Q Developer, and any tool that follows the cross-vendor `AGENTS.md` convention | Auto-loads `AGENTS.md` from the project root. |\n| Cursor | Auto-loads `.cursorrules` (a short stub pointing at `AGENTS.md`). |\n| GitHub Copilot | Auto-loads `.github/copilot-instructions.md` (stub pointing at `AGENTS.md`). |\n| Windsurf | Auto-loads `.windsurfrules` (stub pointing at `AGENTS.md`). |\n| Anthropic Claude Code | Doesn't auto-load `AGENTS.md`. Load it manually with `@AGENTS.md` on the first turn, or add your own per-machine `~/.claude/CLAUDE.md` that references it. The project intentionally does not ship a `CLAUDE.md` mirror. |\n| Google Gemini CLI, JetBrains AI, Replit Agent, anything else | Point the tool at `AGENTS.md` via its config, or load `CONTEXT.md` manually for a shorter orientation. |\n\nIf your tool has a conventional auto-load filename not listed here and you'd like first-class support, open an issue — we'll add a pointer stub.\n\n## Pre-computed indexes\n\n`data/_indexes/` ships 17 derived files so AI consumers can answer cross-reference questions without scanning every skill + catalog. Highlights:\n\n- **`summary-cards.json`** — 100-word abstract per skill; what to load when planning a multi-skill workflow.\n- **`recipes.json`** — 8 curated skill sequences for common use cases (AI red team prep, PCI audit defense, federal IR, DORA TLPT, K-12 EdTech review, ransomware tabletop, new-CVE triage, OSS dep triage).\n- **`chains.json`** — pre-hydrated cross-walks per CVE and per CWE: which skills cite this, which framework gaps it surfaces, which D3FEND countermeasures back it.\n- **`token-budget.json`** — approximate token cost per skill + per section for context budgeting.\n- **`jurisdiction-clocks.json`** — normalized jurisdiction × obligation × hours matrix (breach notification, patch SLA) across 35 jurisdictions.\n- **`did-ladders.json`** — canonical defense-in-depth ladders per attack class (prompt injection, kernel LPE, AI-as-C2, ransomware, supply chain, BOLA, model exfiltration, BEC).\n- **`theater-fingerprints.json`** — structured records for the 7 compliance theater patterns: claim, audit evidence, reality, fast detection test, controls implicated.\n- **`_meta.json`** — sha256 of every source file. The `validate-indexes` predeploy gate fails if any source changed after the last build; `build-indexes --changed` reads this to know what to rebuild.\n\nRegenerate with `exceptd refresh --indexes-only`.\n\n## For skill authors — `agents/`\n\nThe `agents/` directory ships markdown role cards documenting authoring conventions for contributors writing new skills or playbooks. The cards are reference material for humans and AI assistants editing the repo; the CLI runtime does not load them. Operators consuming `@blamejs/exceptd-skills` can ignore the directory.\n\n## Data catalogs\n\nAll skills pull from `data/`. Cross-validated against canonical upstream sources via `exceptd refresh` / `exceptd doctor --cves` / `exceptd doctor --rfcs`.\n\nTo resolve a single citation rather than refresh the whole catalog, `exceptd cve <CVE-ID>` and `exceptd rfc <number>` return a status verdict for one id (catalog → resolved cache → one NVD / datatracker lookup, offline-capable). The lookup caches, so a fan-out of agents shares the answer instead of each independently re-researching the same citation.\n\n- `cve-catalog.json` — CVE metadata with RWEP scores, CISA KEV status, PoC availability, live-patch info\n- `atlas-ttps.json` — MITRE ATLAS v2026.09 TTPs with gap flags and exploitation examples. Each TTP now carries a `cve_refs[]` back-edge — operators reading an ATLAS entry see the catalogued CVEs that cite it without grepping `cve-catalog.json`. The same back-edge is populated on `attack-techniques.json`, and each playbook carries a `_meta.fed_by[]` reverse field naming the upstream playbooks that chain into it.\n- `attack-techniques.json` — MITRE ATT&CK enterprise and ICS techniques, each carrying the catalogued CVEs that cite it in `cve_refs[]`\n- `framework-control-gaps.json` — Per-framework, per-control: what it was designed for vs. what it misses\n- `exploit-availability.json` — PoC locations, weaponization status, AI-assist factor\n- `global-frameworks.json` — All major global compliance frameworks (35 jurisdictions) with control inventories and lag scores\n- `zeroday-lessons.json` — Zero-day → control gap → framework gap → new control requirement mappings\n- `cwe-catalog.json` — CWE entries pinned to CWE v4.20 (Top 25 + AI- / supply-chain-relevant additions)\n- `d3fend-catalog.json` — MITRE D3FEND defensive technique entries pinned to D3FEND v1.3.0\n- `rfc-references.json` — IETF RFC / Internet-Draft references with status, errata, replaces / replaced-by, `last_verified`\n- `dlp-controls.json` — DLP control entries indexed by channel / classifier / surface / enforcement / evidence\n\n---\n\n## Philosophy\n\n**Compliance is not security.** A SOC 2 Type II report confirms that controls existed and operated effectively during the audit period. It says nothing about whether those controls are adequate for current attack patterns. When NIST 800-53 SI-2 says \"apply security patches in a timely manner\" and Copy Fail is a 732-byte deterministic root with a public PoC and no race condition, \"timely\" is the wrong frame entirely.\n\n**Framework lag is measured in months.** MITRE ATLAS v2026.09 (September 2026) is the most current AI threat framework available. It still lags real exploitation by 3-6 months. NIST AI RMF lags by years. ISO 27001:2022 has no AI-specific controls. These skills explicitly flag every place where framework coverage ends and real attacker capability begins.\n\n**AI changed the exploit development timeline.** An AI system discovered Copy Fail in approximately one hour. 41% of 2025 zero-days involved AI-assisted reverse engineering on the attacker side. The time between vulnerability introduction and reliable exploitation is compressing faster than patch management processes can adapt. Risk scoring must reflect this.\n\n**Every org has a compliance theater problem.** The question is not whether paper controls map to audit requirements. The question is whether those controls would actually detect or prevent an attack. These skills answer the second question.\n\n---\n\n## Contributing\n\nSee [CONTRIBUTING.md](CONTRIBUTING.md). Key rules:\n\n- No new CVE reference without a complete `data/cve-catalog.json` entry\n- No new framework gap claim without a `data/framework-control-gaps.json` entry\n- No skill uses CVSS as the sole risk metric\n- Every new zero-day triggers a `data/zeroday-lessons.json` entry\n\n---\n\n## License\n\nApache 2.0. See [LICENSE](LICENSE).\n\nCommunity at [exceptd.com](https://exceptd.com).\n","readmeFilename":"README.md"}
![\"exceptd\"](\"https://raw.githubusercontent.com/blamejs/exceptd-skills/main/public/img/logo/exceptd-logo-dark.svg\"){kind=logo}
\n