Audit installed AI agent skills against the OWASP Agentic Skills Top 10 and emit text, JSON, SARIF, or HTML reports.
Memory
compliance-aiops
Try itUse this skill whenever the user needs compliance evidence from the audit trails their governed AIops agents already write — mapping AI-agent infra-ops activity to HIPAA §164.312, PCI-DSS v4.0, SOC 2 TSC, or GDPR controls, producing a change-approval report, a gap analysis, an exceptions/anomaly report, or a hash-chain-sealed, tamper-evident evidence bundle. Always use this skill for "compliance evidence", "HIPAA / PCI-DSS / SOC 2 / GDPR evidence", "audit trail report", "coverage for control X", "which controls are we short on / gap analysis", "who approved this change / change-management evidence", "denied or errored ops / anomaly evidence", "seal / sign an evidence bundle", "prove this bundle wasn't altered", or "detect deleted audit rows". Do NOT use to scan or operate infrastructure and do NOT treat it as a GRC platform — it reads the local audit databases the OTHER AIops-tools write and converts them to evidence; for platform operations use those other AIops-tools. Evidence, not c
What it does
Use this skill whenever the user needs compliance evidence from the audit trails their governed AIops agents already write — mapping AI-agent infra-ops activity to HIPAA §164.312, PCI-DSS v4.0, SOC 2 TSC, or GDPR controls, producing a change-approval report, a gap analysis, an exceptions/anomaly report, or a hash-chain-sealed, tamper-evident evidence bundle. Always use this skill for "compliance evidence", "HIPAA / PCI-DSS / SOC 2 / GDPR evidence", "audit trail report", "coverage for control X", "which controls are we short on / gap analysis", "who approved this change / change-management evidence", "denied or errored ops / anomaly evidence", "seal / sign an evidence bundle", "prove this bundle wasn't altered", or "detect deleted audit rows". Do NOT use to scan or operate infrastructure and do NOT treat it as a GRC platform — it reads the local audit databases the OTHER AIops-tools write and converts them to evidence; for platform operations use those other AIops-tools. Evidence, not certification. Reads sibling audit trails read-only; no external API, no network, no platform credentials. Fully offline and deterministic.
The skill document
Compliance AIops
Disclaimer: Community-maintained open-source project, not affiliated with, endorsed by, or sponsored by any framework body or GRC vendor. HIPAA, PCI-DSS, SOC 2, GDPR and OSCAL are referenced descriptively; trademarks belong to their owners. Source at github.com/AIops-tools/Compliance-AIops under the MIT license.
Governed compliance-evidence tooling — 19 MCP tools. It reads the audit
trails your governed AIops agents already write (~/.-aiops/audit.db, one
shared audit_log schema, discovered via ~/.*-aiops/audit.db) read-only,
and turns that activity into framework-mapped, hash-chain-sealed compliance
evidence. It does not scan infrastructure and does not replace a GRC
platform.
Standalone: the governance harness is bundled (
compliance_aiops.governance). Not a platform wrapper — no external API, no network, no platform credentials. Evidence, not certification; fully offline and deterministic.
What This Skill Does
| Group | Tools | Count | Read/Write |
|---|---|---|---|
| Audit reads | list_audit_sources, query_audit_events, activity_timeline | 3 | read |
| Framework mapping | list_frameworks, coverage_summary, control_evidence, gap_analysis | 4 | read |
| Assurance reports | approval_report, exceptions_report | 2 | read |
| Integrity | verify_source_chain, verify_bundle, list_bundles, bundle_schedule_hint, oscal_assessment_results | 5 | read |
| Artifacts | generate_evidence_bundle (low), export_bundle (low), sign_bundle (medium) | 3 | write (no external mutation) |
| Undo | undo_list, undo_apply | 2 | undo |
Frameworks & sample controls
| Framework | Sample controls (strength) |
|---|---|
| HIPAA §164.312 | 164.312(b) Audit controls (strong), 164.312(a)(1) Access control (strong), 164.312(c)(1) Integrity (strong) |
| PCI-DSS v4.0 | 10.2 Audit log content (strong), 10.3 Protect audit logs (strong), 7-8 Least privilege / authn (partial) |
| SOC 2 TSC | CC6.1 Logical access (strong), CC7.2 Monitoring (strong), CC8.1 Change management (strong) |
| GDPR | Art.30 Records of processing (partial), Art.32 Security of processing (strong) |
| ISO/IEC 27001:2022 (Annex A) | A.5.15 Access control (strong), A.5.16 Identity mgmt (strong), A.5.18 Access rights (partial), A.8.2 Privileged access (partial), A.8.15 Logging (strong), A.8.16 Monitoring (strong), A.8.32 Change management (strong) |
| 等保2.0 (DJCP L3) GB/T 22239-2019 三级 | 8.1.5.4 安全审计 (strong), 8.1.4.2 访问控制 (partial), 8.1.5 安全管理中心/集中审计 (strong) |
Audit trails prove operating effectiveness strongly but control design /
configuration only partially — each control is labelled strong or partial,
and gap_analysis surfaces the caveat rather than overclaiming.
Quick Install
uv tool install compliance-aiops
compliance-aiops init # discover sibling ~/.*-aiops/audit.db, set org name, optional signing key
compliance-aiops doctor # which sibling audit DBs are present/readable
When to Use This Skill
- Map AI-agent infra-ops activity to a framework's controls (
coverage_summary) - Pull the evidence rows + population for one control (
control_evidence) - Find controls with no or weak evidence, with the honest caveat (
gap_analysis) - Produce a change-approval artifact — who approved which high-risk write and why
(
approval_report) - Produce enforcement / anomaly evidence — denied / errored / budget-tripped ops
(
exceptions_report) - Seal a tamper-evident evidence bundle (
generate_evidence_bundle,sign_bundle) and later prove it wasn't altered (verify_bundle) - Detect deleted / missing audit rows in a source trail (
verify_source_chain)
Do NOT use to scan or operate infrastructure, or as a GRC platform. It reads the audit DBs the other AIops-tools write; for platform operations use those other AIops-tools.
Related Skills — Skill Routing
| If the user wants… | Use |
|---|---|
| Compliance evidence from existing AIops audit trails | compliance-aiops (this skill) |
| To actually operate a platform (VMs, storage, clusters, network, …) | the relevant platform AIops-tools skill |
| OT / industrial edge (Modbus, OPC-UA, PLC) | the industrial-aiops line |
| A full GRC platform / policy management | out of scope — this is evidence, not GRC |
Common Workflows
1. "The SOC 2 auditor wants Q3 change-approval evidence by Friday"
compliance-aiops doctor→ confirm the source audit trails are discoverable and readable before you promise a delivery datecompliance-aiops report sources(MCP:list_audit_sources) → which sibling audit trails were found, and the event count and date range in each. If a source you expected is missing, the bundle would be silently incomplete — fix discovery firstcompliance-aiops report coverage soc2→ confirm CC8.1 is actually covered by the evidence you have, before generating anythingcompliance-aiops report approvals→ the high-risk write operations with their named approver and rationale — this is the population CC8.1 is asking aboutcompliance-aiops bundle generate soc2 --since 2026-07-01 --until 2026-10-01 --sign→ a hash-chain-sealed bundle under~/.compliance-aiops/bundles/compliance-aiops bundle export --format markdown→ the auditor-facing report (alsojson/csv)- Failure branch: if
report coverageshows CC8.1 thin, do not generate anyway and hope — run workflow 2 first and hand the auditor the honest gap statement. A bundle asserts what the audit trail contains; it cannot manufacture evidence that was never recorded.
2. "Which controls are we actually short on?" (gap analysis)
compliance-aiops report sources→ establish the evidence base and its date coverage; a gap caused by a missing source is a different problem from a gap caused by missing activitycompliance-aiops report gaps hipaa(alsopci_dss,soc2,gdpr) → controls with no or weak evidence, each with an honest caveat and a remediation suggestioncompliance-aiops report exceptions→ the operations that ran without an approver or rationale — usually the fastest-to-fix category of gap- Drill into one control's population with
control_evidence(MCP) to see the reproducible query behind the coverage number, so the figure can be defended rather than merely quoted compliance-aiops report coverageagain after remediation to confirm the gap actually closed- Failure branch: if
list_frameworksdoes not carry the framework or control the auditor named, say so — this tool maps to HIPAA / PCI-DSS / SOC 2 / GDPR and does not silently substitute a near-miss control.
3. Prove a delivered bundle was not altered
compliance-aiops bundle list→ locate the bundle and its recordedchainHeadcompliance-aiops bundle verify→ re-derives the hash chain, compares it to the seal'schainHead, and checks the optional signature- Because the chain is computed over evidence records only, the same
(framework, period, sources) reproduces the same
chainHead— regenerate and compare to prove reproducibility - Record the
chainHeadout-of-band (ticket, email to the auditor, WORM store) at delivery time; that out-of-band copy is what makes later verification meaningful verify_source_chain(MCP) on each source → returns the source chain head and flags row-id gaps, a sign that rows were deleted from thataudit.db- Failure branch: a
chainHeadmismatch or a row-id gap means the evidence is not trustworthy — escalate, and treat the sourceaudit.dbas the system of record. Do not re-seal a fresh bundle to make the mismatch go away; the tool is tamper-evident, not tamper-proof, and its whole value is that it reports this rather than papering over it.
4. 定期封存 — schedule periodic sealed bundles (no daemon)
This tool ships no scheduler; it emits a cron line for you to install.
compliance-aiops report sources→ confirm the sources you want sealed are discoverable from the account cron will run as (a common failure: cron sees a different$HOME)compliance-aiops bundle schedule soc2 --cron "0 2 * * 1" --period 7d --sign(MCP:bundle_schedule_hint) → returns acronLineplus the exact non-interactive command. It writes nothing.- Paste the
cronLineintocrontab -e, e.g.0 2 * * 1 compliance-aiops bundle generate soc2 --period 7d --sign - Export
COMPLIANCE_AIOPS_MASTER_PASSWORDin the cron environment so the signing key unlocks non-interactively — never inline the real password in the crontab - After the first scheduled run,
compliance-aiops bundle listandbundle verifythe newest bundle to confirm the unattended path really works - Failure branch: if the cron run produces no bundle, the usual causes are
an unset master password (signing cannot unlock) or
COMPLIANCE_AIOPS_HOMEnot being set in cron's environment, so sources resolve elsewhere. Verify by running the emitted command by hand with a clean environment before trusting the schedule.
Governance & Safety
The skill reads audit trails and writes evidence bundles and records what it does; it does not decide whether producing or signing a bundle is permitted. That is your agent's judgement, or the filesystem permissions of the account it runs as. There is no read-only switch, policy file, or approval gate.
- Audit is the guarantee, and it is not bypassable. Every operation — MCP and CLI alike — is logged to
~/.compliance-aiops/audit.db(relocatable viaCOMPLIANCE_AIOPS_HOME): params, result, status, duration, and the risk tier. The CLI writes the same row the MCP path does. - The source audit trails are opened read-only; the tool never mutates them. The only files written are bundles under
~/.compliance-aiops/bundles/. COMPLIANCE_AUDIT_APPROVED_BY/COMPLIANCE_AUDIT_RATIONALEare optional annotations recorded on the audit row (who/why); they are never required and never block.- Tamper-EVIDENT, not tamper-PROOF — the source
audit.dbremains the system of record.
References
references/capabilities.md— full tool → inputs → returns referencereferences/cli-reference.md— CLI command referencereferences/setup-guide.md— source discovery, org name, optional signing key, integrity notes
Related skills
Use this skill whenever the user needs to observe or govern on-endpoint local LLMs running on Ollama, llama.cpp (llama-server), LM Studio, or a local single-node vLLM — inventory installed/running models with an allow/deny verdict (shadow-AI detection), inspect VRAM residency, model license/params/capabilities and server version, view the model policy, detect model provenance/digest drift (re-pulled or tampered weights; strong for Ollama/llama.cpp, id-only and honestly weaker for LM Studio/vLLM), scan a prompt for secrets / PII / source-code / jailbreak with a weighted risk band, route a prompt THROUGH a guard that scans + policy-gates + records + runs-if-allowed (guarded_generate / observe_chat), query the observed-usage log, and roll up anomalies (shadow models, digest drift, high-risk + blocked prompts). Always use this skill for "what local models are installed", "find shadow / unsanctioned AI models", "which model is loaded in VRAM", "scan this prompt for secrets/PII before sendin
Security and compliance auditing tool for AI agents. Scans code for vulnerabilities, checks GDPR/CCPA compliance, generates risk reports with remediation guidance.
Audit a principal AI agent or coordinator bot: review memory, learnings, recent errors, installed skills, operational risks, delegation posture, and propose...
A guided HIPAA compliance posture assessment that ends in a shareable Word report.
Use the Cryptowerk Verifiable Data skill to implement an "Agent Flight Recorder" that creates an immutable, cryptographically verifiable audit trail for Open...