How do I install the System Audit skill?

Run npx skillmds@latest add onlinesourdough/system-audit in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is System Audit safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with System Audit?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is System Audit free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published System Audit?

onlinesourdough (@onlinesourdough) published this skill. Their other Agent Skills are listed on their SkillMD profile.

System Audit by onlinesourdough — Agent Skill

Name: System Audit
Author: onlinesourdough

File contents

Periodic System audit

Use this route for an accumulated-state health check, not for setup, repair, release work, per-change Review, or a single Power BI deliverable.

Input

Select exactly one scope:

repository: Git identity, clean state, live upstream relation, specialist routes, and deterministic validation surface;
workspace: project truth, run/history/proof relations, and discoverable failure/recovery evidence;
both: both surfaces plus the existing deterministic structural checks.

Run from the exact repository root:

python3 workspace/engine/system_audit.py --scope repository
python3 workspace/engine/system_audit.py --scope workspace
python3 workspace/engine/system_audit.py --scope both

Output

Return exactly one status: PASS, FAIL, or BLOCKED, followed by observable evidence, evidence gaps, and the smallest next action. PASS requires all selected evidence and no observed defect. FAIL means a defect is observable. BLOCKED means required evidence is unavailable and no defect already proves failure. Uncertainty is never PASS.

Read-only boundary

Do not fetch into the audited repository, repair a finding, generate a run, append the ledger, promote an example, create an issue, publish, change a service, or alter any audited ref, index, worktree, or workspace file. The reference route reads the live upstream with ls-remote and loads commits only into temporary storage for ancestry comparison. It records no audit run. Reject embedded remote credentials before any Git network command; credentials belong in an external helper or SSH agent, never in the configured remote URL.

Use the powerbi audit contract by reference when a bounded business/domain audit is needed. Use validation by reference for deterministic PBIP/PBIR/TMDL proof. Semantic or visual judgment remains a specialist review, and per-change Review remains the lifecycle gate for proposed changes; neither is this periodic audit.

Route a System finding to its owning Build and per-change Review. If the work originated in an AIOS improvement flow and the finding concerns that flow rather than this repository's local truth, route it back to AIOS improvement triage. The audit never performs the repair or authorizes Ship.

---name: system-auditdescription: Run a strictly read-only periodic audit of accumulated Agentic Power BI System repository or workspace health.---# Periodic System auditUse this route for an accumulated-state health check, not for setup, repair,release work, per-change Review, or a single Power BI deliverable.## InputSelect exactly one scope:- `repository`: Git identity, clean state, live upstream relation, specialist routes, and deterministic validation surface;- `workspace`: project truth, run/history/proof relations, and discoverable failure/recovery evidence;- `both`: both surfaces plus the existing deterministic structural checks.Run from the exact repository root:```shpython3 workspace/engine/system_audit.py --scope repositorypython3 workspace/engine/system_audit.py --scope workspacepython3 workspace/engine/system_audit.py --scope both```## OutputReturn exactly one status: `PASS`, `FAIL`, or `BLOCKED`, followed by observableevidence, evidence gaps, and the smallest next action. `PASS` requires allselected evidence and no observed defect. `FAIL` means a defect is observable.`BLOCKED` means required evidence is unavailable and no defect already provesfailure. Uncertainty is never `PASS`.## Read-only boundaryDo not fetch into the audited repository, repair a finding, generate a run,append the ledger, promote an example, create an issue, publish, change aservice, or alter any audited ref, index, worktree, or workspace file. Thereference route reads the live upstream with `ls-remote` and loads commits onlyinto temporary storage for ancestry comparison. It records no audit run.Reject embedded remote credentials before any Git network command; credentialsbelong in an external helper or SSH agent, never in the configured remote URL.Use the `powerbi` audit contract by reference when a bounded business/domainaudit is needed. Use `validation` by reference for deterministic PBIP/PBIR/TMDLproof. Semantic or visual judgment remains a specialist review, and per-changeReview remains the lifecycle gate for proposed changes; neither is thisperiodic audit.Route a System finding to its owning Build and per-change Review. If the workoriginated in an AIOS improvement flow and the finding concerns that flowrather than this repository's local truth, route it back to AIOS improvementtriage. The audit never performs the repair or authorizes Ship.

Frequently asked questions

System Audit