AgentStack
SKILL verified MIT Self-run

Audit Gnadd

skill-alexhagemeister-gnadd-audit-gnadd · by AlexHagemeister

>-

No reviews yet
0 installs
1 views
0.0% view→install

Install

$ agentstack add skill-alexhagemeister-gnadd-audit-gnadd

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README — it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-alexhagemeister-gnadd-audit-gnadd)

Reliability & compatibility

Security review passed
0 installs to date
no reviews yet
5d ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming — see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps — measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Audit Gnadd? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

GNADD Audit

Review a repository against GNADD workflow principles and propose a minimal set of alignment fixes. This is read-only: do not create, edit, stage, commit, stash, or delete files; do not auto-create GitHub issues.

Auto-Invocation Gate

If this skill was auto-selected from context rather than explicitly invoked with /audit-gnadd, stop before running commands or reading repo context. Briefly explain why a GNADD audit appears useful and ask: "Run /audit-gnadd now?" Proceed only after confirmation.

GNADD Invariants

  • GitHub is the system of record; tracking state belongs in issues, branches, PRs,

and commits — not in markdown files.

  • Context-file scrutiny is the primary audit dimension; git/workflow checks are

secondary alignment signals.

  • The audit complements prime-gnadd (live snapshot) — it does not replace session

orientation or duplicate its narrative output.

  • Proposed fixes are slices for the user to capture via /new-issue-gnadd; the agent

never files issues autonomously.

  • For broader workflow rationale, use help-gnadd.

Purpose

Answer: Is this repo aligned with GNADD principles, and what is the smallest fix set to get there?

Build a structured report covering:

  • Context files that violate or risk violating the describe-vs-track rule
  • Shallow git/workflow misalignment (orphaned branches, diverged main, stashes)
  • Whether GNADD skills appear installed
  • Minimal proposed fixes, each mappable to one /new-issue-gnadd

Do not read source code, schemas, tests, or dependencies. Do not perform deep commit archaeology.

Workflow

1. Load Audit Criteria

Before inspecting the repo, load GNADD file-hygiene and workflow principles:

  1. Use the describe-vs-track rule from help-gnadd if already in context.
  2. Fetch Part 1 of the canonical guide from this pinned URL (same pin as

help-gnadd — do not use main):

https://raw.githubusercontent.com/AlexHagemeister/gnadd/v0.3.0/GNADD.md

(The pin is rewritten by scripts/release.sh at release time.)

  1. If the repo being audited contains a local GNADD.md, reading Part 1 locally

is an acceptable fallback or supplement.

Extract the operational taxonomy:

| Category | Examples | Default severity | |---|---|---| | Tracking files | tasks.md, TODO.md, progress/session-state files, maintained plans, status checklists | Violation | | Expired phase artifacts | requirements.md, standalone PRD/spec not distilled into README + issues | Warning | | Misplaced decisions | decisions.md, ADR folders used as live tracking | Warning | | Describe-only (OK) | README vision, AGENTS.md, agent rules, archived/dated requirements | OK | | Missing describe content | README without project-level "What done looks like" on a GNADD-shaped project | Warning |

Apply the tiebreaker from GNADD Part 1:

> Would the agent need to keep this file up to date? If yes → tracking violation.

2. Discover Context Files

From the repo root, run a shallow filename discovery (depth-capped, skip .git, node_modules, vendor, dist, build, .next, .venv, coverage, target):

find . -maxdepth 4 -type f \( \
  -iname 'tasks.md' -o -iname 'task.md' -o -iname 'todo.md' -o -iname 'todo.txt' \
  -o -iname 'requirements.md' -o -iname 'prd.md' -o -iname 'spec.md' \
  -o -iname 'decisions.md' -o -iname 'progress.md' -o -iname 'status.md' \
  -o -iname 'roadmap.md' -o -iname 'changelog-dev.md' \
  -o -iname 'session.md' -o -iname 'session-state.md' \
\) -not -path './.git/*' -not -path '*/node_modules/*' -not -path '*/vendor/*' \
  -not -path '*/dist/*' -not -path '*/build/*' -not -path '*/.next/*' \
  -not -path '*/.venv/*' -not -path '*/coverage/*' -not -path '*/target/*' 2>/dev/null

Always read README.md at the repo root if it exists.

For each discovered file (and README.md):

  • Read enough to classify — usually the first ~50 lines suffices.
  • Look for tracking signals: - [ ] / - [x] checkboxes, "Status:",

"In Progress", sprint/progress sections, satisfaction tracking, maintained task lists, "current focus" notes the agent would need to update.

  • Look for safe signals: archive headers, "as of DATE", explicit notes that

content was distilled into issues/README, pure vision/convention text with no mutable state.

Classify each file as Violation, Warning, or OK with a one-line rationale. When filename and content conflict, content wins.

3. Check GNADD Skills Presence

Determine whether GNADD skills appear available to the agent:

ls .agents/skills/ 2>/dev/null

Look for help-gnadd, prime-gnadd, new-issue-gnadd, or similar GNADD skill names. Optionally try npx skills list 2>/dev/null if the skills CLI is available.

If undetectable or absent, report as Info — not a blocker. Include the minimal install recommendation:

npx skills add AlexHagemeister/gnadd -g -a  --copy -y

The file and git audit still runs regardless.

4. Shallow Git / Workflow Alignment

Verify GitHub CLI when checking workflow alignment:

gh auth status

If gh is not installed or authenticated, warn the user, skip GitHub cross-checks, and still report local git signals from the commands below.

When auth is confirmed (or for local-only checks), run from the repo root:

git fetch --prune origin 2>/dev/null || true
git branch --show-current
git branch --list
git stash list
git log --oneline origin/main..main 2>/dev/null
git log --oneline main..origin/main 2>/dev/null
gh issue list --state open 2>/dev/null
gh pr list --state open 2>/dev/null

git fetch --prune updates remote-tracking refs only — read-only, same as prime-gnadd. If no remote is configured, note it and continue with local signals.

Interpret for alignment only — do not produce a session-orientation narrative:

| Signal | Severity | Finding | |---|---|---| | origin/main..main shows commits | Violation | Local main ahead of origin — direct commits bypassing the loop | | main..origin/main shows commits | Info | Local main behind origin — sync before adopting the loop | | Non-empty stash list | Warning | Invisible WIP outside issues/branches | | Branch issue-/* with no open issue #N | Warning | Orphaned issue branch | | Open issue #N with no branch and no recent merged PR | Info | Issue may be unstarted | | Active WIP branch not matching issue-/ | Warning | Off-loop branch naming | | Open PR with no issue reference in title/body | Warning | Weak issue linkage |

Do not review individual commits, run blame, or lint commit messages.

Cross-reference open issues with local branches by issue number extracted from branch names (issue-/...).

5. Synthesize Minimal Fixes

For every Violation and Warning, propose one concrete minimal fix — not a broad rewrite plan. Group related fixes into proposed issue slices the user can capture separately via /new-issue-gnadd.

Examples:

  • tasks.md exists → "Migrate open items to GitHub issues, then delete the

file."

  • Stale requirements.md → "Archive with a date header or delete after

distilling stable conclusions into README and actionable items into issues."

  • Standalone PRD with tracking content → "Collapse vision into README 'What

done looks like'; move actionable specifics to issues."

  • GNADD skills not detected → "Install GNADD skills globally per README."
  • Diverged local main → "Resolve main divergence before starting new issue

work — the sanctioned path is gnadd.sh doctor --rescue-main (bundled with the prime-gnadd skill), which is lossless and never resets."

  • No branch ruleset / merge policy on the repo → "Run gnadd.sh init once to

enforce squash-only merges and PR-required main at the server."

Keep the fix set minimal — prefer the smallest change that restores alignment.

6. Produce the Report

Use the output format below. Lead with violations; do not bury context-file findings below git noise.

Classification Reference

Violation — Confirmed tracking file or workflow habit that actively undermines GNADD (task lists, progress files, diverged main with local-only commits).

Warning — Likely misalignment needing human judgment (expired requirements.md, fat PRD, orphaned branch, missing README criteria section).

OK — Describe-only content correctly structured (vision docs, AGENTS.md, archived phase artifacts clearly marked).

Info — Informational, not necessarily a problem (skills install unverified, issue with no branch yet, main behind origin).

Output Format

## Summary

## Context Files
### Violations
- `` — . **Fix:** 

### Warnings
- `` — . **Fix:** 

### OK
- `` — 

## Workflow Alignment

## GNADD Skills

## Proposed Fixes
1. 
2. ...

## Already Aligned

Keep the report scannable. Each proposed fix should be actionable without reading the rest of the audit.

Closing Guidance

Offer a brief next-step nudge only when the audit completed successfully — not when gh auth failed mid-run and GitHub checks were skipped.

Skip when the audit halted early or critical checks could not run — tell the user what's missing first.

Infer one primary suggestion from findings, plus at least one alternative when ambiguous:

  1. Violations present → nudge /new-issue-gnadd for the highest-severity fix;

offer capturing the full proposed-fix list as separate issues.

  1. Warnings only, no violations → nudge confirming which warnings are

intentional before filing issues; offer /new-issue-gnadd for any the user wants to act on.

  1. Clean audit → nudge /prime-gnadd for ongoing session work, or /new-issue-gnadd

if new work is ready to capture.

  1. GNADD skills not detected on a repo being adopted → nudge installing

skills first, then re-running /audit-gnadd after remediation issues are filed.

Keep it to a sentence or two with invitational options. Do not restate the full GNADD workflow. Never auto-create issues.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.