aim-audit
Evolving product-truth audit: one living artifacts/AUDIT_<SLUG>.md, update in place (aim-handoff rules), never dated _REAUDIT/_AUDIT3 novels.
Skill by BrianV1981
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/brianv1981-aim-audit
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: security
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/aim-audit
What's inside
- Skill name
aim-audit- Description
- Evolving product-truth audit: one living artifacts/AUDIT_<SLUG>.md, update in place (aim-handoff rules), never dated _REAUDIT/_AUDIT3 novels. Modes baseline|delta|closeout|event. Use for audit, re-audit, /aim-audit, or when the Operator asks how a target is doing vs its README.
- License
- MIT
- Instructions
- 135 lines in SKILL.md · view on GitHub
Files (2)
- SKILL.md
- PROTOCOL.md · references/
SKILL.md (first lines)
# `aim-audit` > **MANDATE:** Product-truth, not a new literary work. If the living audit exists, **amend it**. Pair with persona **`aim-auditor`** (aim-coagents) when you need the hat. This skill is the **procedure**. Canonical rubric: [`references/PROTOCOL.md`](references/PROTOCOL.md). Read it on **baseline**, **event**, or when grades/IDs are unclear. Do not inline it into `AGENTS.md`. --- ## 1. Law (same spirit as `aim-handoff`) 1. **UPDATE, do not decapitate.** One file: `artifacts/AUDIT_<SLUG>.md` (slug `A-Z0-9_`, no date). 2. **No sibling novels.** `_REAUDIT`, `_AUDIT3`, `_<YYYY-MM-DD>` are legacy snapshots, not the working copy. 3. **No guesses.** Unknown → `UNKNOWN`. Proven → path, SHA, issue #, command, exit code, count. 4. **Target repo stays clean** unless the Operator scoped a fix. You write the audit file on *this* desk. 5. **Default mode is `delta`** when the living file exists. ## 2. Modes | Mode | When | Write | |------|------|--------| | **baseline** | No living file, or Operator says full E2E from zero | Full §0–§10. Invent smoke contract. Mint finding IDs. | | **delta** | **Default** later pass | Re-run smoke. Update statuses. Rewrite §2 + §9. Append §10 row. Do not rewrite identity/architecture unless they changed. | | **closeout** | Series should stop | Freeze vs backlog. Event-driven from here. | | **event** | Scope change or incident | Snapshot first if sharing. May rewrite §1 / §8. | ## 3. Checklist (every pass) ```text 1. Slug + clone + tip SHA. 2. Living file missing → baseline; else delta (unless Operator said closeout/event). 3. Read the living file. Verify §1; do not re-derive from README if it exists. 4. git log <prior_sha>..HEAD — material? empty close? 5. Re-run §8 smoke with timeouts. Fill §3 Last/Now. 6. Open IDs: close / partial / regress. Refute new critical/high. 7. Mint IDs only for new defects (<SLUG>-NNN). 8. Rewrite §2 verdict + §9 queue. Update changed §6 rows. §7 scan. 9. Append §10 row. Bump Pass N. Header SHA/grade/date. 10. Stop. Do not file GH issues on the target unless asked. ``` ## 4. Template (roles stay; numbers stay) ```markdown # Audit: <product> | Field | Value | |-------|--------| | **Repo** | url | | **Clone** | path | | **Commit** | `sha` — subject (stamp) | | **Mode** | baseline \| delta \| closeout \| event | | **Pass** | N | | **Overall grade** | letter (was prior) | | **Auditor** | vessel / persona aim-auditor | | **Updated** | stamp | ## 0. CLOSED FINDINGS (DO NOT RE-HUNT) ## 1. IDENTITY ## 2. VERDICT ## 3. SMOKE MATRIX ## 4. OPEN FINDINGS ## 5. DELTA THIS PASS ## 6. GRADE ROLLUP ## 7. SECURITY ## 8. ARCHITECTURE / CONTRACT ## 9. NEXT ACTIONS ## 10. PASS LOG ``` Delta: §5 is the only new narrative. §10 appends one row. Move closed IDs to §0. ## 5. Findings `STATUS`: `open` | `partial` | `closed` | `wontfix` | `regressed` `SEV`: `critical` | `high` | `med` | `low` Close only with this-pass evidence. Empty `Closes #N` (`git show --name-only` blank) is a **process-integrity** finding. Regress → move ID back from §0 to §4; do not mint a new ID. **Refute** critical/high before filing. ## 6. Grades (short) | Letter | Meaning | |--------|---------| | **F** | Advertised happy path cannot complete | | **C** | Path works; docs/CI/board lie or tests absent | | **B** | Daily-driver; residual polish | | **A** | Public-OSS: tests exercise the path; docs match CLI | `overall = min(secrets area, worst *advertised* critical path)`. Process theater can cap at **C**. Full rubric: PROTOCOL. ## 7. Smoke (minimum; specialize in §8) Use timeouts on embed/network paths. Prefer the target’s own `test.sh` / pytest when present. A.I.M. vessels (adapt names; do not assume a global `aim` alias): ```bash ./aim --help ./aim doctor ./aim map timeout 25 ./aim search worktree --top-k 1 # promote: print repo_root; test -d "$repo_root/.git" # LICENSE: gh api repos/<owner>/<repo>/license --jq .license.spdx_id # secrets heuristic on tracked files only ``` Host CLIs differ; invoke the project wrapper or `joshua_os/.aim_core/aim_cli.py` (legacy nest: `aim-agy_os/`). ## 8. File / edit tools (dual vessel) When creating or amending the living audit: | Host | Create / overwrite | Edit in place | Read | |------|--------------------|---------------|------| | **Grok** | `write` | `search_replace` | `read_file` | | **AGY** | `write_to_file` | `replace_file_content` | host read |
From BrianV1981/aim-skill-library, MIT.
More from BrianV1981/aim-skill-library
- 0.601score
aim-memory-wiki
BrianV1981
Interactive active-memory ingestion skill for maintaining a persistent, compounding LLM knowledge base (the Memory Wiki).
SkillAFreshness: A★ 3Claude Code +9 - 0.601score
aim-handoff
BrianV1981
Ephemeral context handoff: structure Engineering Handoff docs to the gold-standard numbered template (HANDOFF style), update in place, optional Tmux baton…
SkillAFreshness: A★ 3Claude Code +9 - 0.601score
aim-gitops
BrianV1981
Automates the strict GitOps worktree and TDD workflow (formerly `aim fix` and `aim promote`) natively using PowerShell and git.
SkillAFreshness: A★ 3Claude Code +9 - 0.601score
aim-mega-guide
BrianV1981
A highly specialized content pipeline for generating consistent, deeply structured 'Mega Guides' and SOPs from a single, long-form YouTube video or podcast…
SkillAFreshness: A★ 3Claude Code +9 - 0.597score
aim-mega-article
BrianV1981
Multi-source B2B authority article pipeline combining 3-5 contextual YouTube masterclass embeds, inline hyperlinked empirical data citations, 3 bespoke…
SkillAFreshness: A★ 3Claude Code +9 - 0.569score
aim-communicate
BrianV1981
Inter-agent communication via tmux with permission gates, chalkboard + short-paste protocol (Grok/AGY-hardened).
SkillBFreshness: B★ 3Claude Code +9 - 0.566score
aim-calc
BrianV1981
Stateful scientific calculator for agents.
SkillBFreshness: B★ 3Claude Code +9 - 0.564score
aim-export-cartridge
BrianV1981
Export A.I.M.
SkillBFreshness: B★ 3Claude Code +9 - 0.564score
aim-grok-context
BrianV1981
Project context for the aim-grok vessel (A.I.M.
SkillBFreshness: B★ 3Claude Code +9 - 0.564score
aim-bwrap-forge
BrianV1981
Dynamically spin up bubblewrap (bwrap) isolated sandbox directories and optional tmux co-agent sessions for untrusted or multi-tenant work.
SkillBFreshness: B★ 3Claude Code +9 - 0.558score
aim-projects
BrianV1981
Interacts with the GitHub Projects kanban board natively using the `gh` CLI on Windows.
SkillBFreshness: B★ 3Claude Code +9 - 0.558score
aim-google
BrianV1981
Fast, script-friendly CLI for interacting with Google Workspace (Gmail, Calendar, Drive, Docs, etc.)
SkillBFreshness: B★ 3Claude Code +9
Signals
- Stars
- 3
- Forks
- 0
- Open issues
- 6
- Last commit (this folder)
- Aug 12, 2026 (2 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/BrianV1981/aim-skill-library
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from BrianV1981
- 0.601score
aim-gitops
BrianV1981
Automates the strict GitOps worktree and TDD workflow (formerly `aim fix` and `aim promote`) natively using PowerShell and git.
SkillAFreshness: A★ 3Claude Code +9 - 0.601score
aim-mega-guide
BrianV1981
A highly specialized content pipeline for generating consistent, deeply structured 'Mega Guides' and SOPs from a single, long-form YouTube video or podcast…
SkillAFreshness: A★ 3Claude Code +9 - 0.601score
aim-memory-wiki
BrianV1981
Interactive active-memory ingestion skill for maintaining a persistent, compounding LLM knowledge base (the Memory Wiki).
SkillAFreshness: A★ 3Claude Code +9 - 0.601score
aim-handoff
BrianV1981
Ephemeral context handoff: structure Engineering Handoff docs to the gold-standard numbered template (HANDOFF style), update in place, optional Tmux baton…
SkillAFreshness: A★ 3Claude Code +9 - 0.597score
aim-mega-article
BrianV1981
Multi-source B2B authority article pipeline combining 3-5 contextual YouTube masterclass embeds, inline hyperlinked empirical data citations, 3 bespoke…
SkillAFreshness: A★ 3Claude Code +9 - 0.569score
aim-communicate
BrianV1981
Inter-agent communication via tmux with permission gates, chalkboard + short-paste protocol (Grok/AGY-hardened).
SkillBFreshness: B★ 3Claude Code +9
Cite this page
aim-audit skill. Agent Mart. https://agentmart.dev/skills/brianv1981-aim-audit (last updated 2026-10-24)