Skip to main content
Agent skill

Advisory

File, maintain, and close security advisories. Keeps docs/security/advisories/, docs/security/practices.md, and GitHub Security Advisories (GHSA) in sync. Two modes — new (file advisory) and close (mark fixed).

Get started

Get the complete skill folder, including its scripts and reference files.

When to use this skill

Use Advisory when an AI agent needs a reusable SKILL.md workflow for this job: File, maintain, and close security advisories. Keeps docs/security/advisories/, docs/security/practices.md, and GitHub Security Advisories (GHSA) in sync. Two modes — new (file advisory) and close (mark fixed).

When not to use it

Skip Advisory when the task is outside the coding category, or when a more specific skill in this directory already covers the same workflow with clearer triggers.

How to install

Where will you use it?

Install scope
Claude Code folder setup
mkdir -p ~/.claude/skills/advisory
# Save SKILL.md to ~/.claude/skills/advisory/SKILL.md
  1. Claude Code personal install: place the skill and its bundled files at ~/.claude/skills/advisory/SKILL.md. For a shared project, commit the folder at .claude/skills/advisory/ instead.
  2. Current Claude Code detects SKILL.md changes in watched directories during the session. If you just created a top-level skills directory, run /reload-skills. Check that the skill is listed before testing its trigger.
  3. For Claude chat or Cowork in the Desktop app, manage uploaded skills in Customize > Skills; copying files into a local Claude Code folder is not a Desktop install. Other agents may use different install locations.

Full install guide for Claude, Cursor, and Codex →

Try a starter prompt

After installing the skill, paste this starter prompt into your agent to verify it triggers on a small task.

Starter prompt
Use the Advisory skill for this task: A vulnerability has been identified and needs a disclosure record. Confirm any required inputs or files before making changes.

What supports this recommendation?

80/100 provisional content score. This measures documentation, practical detail and source signals. It is not a measured task success rate or a security certification.

Documentation
87/100
Practical detail
92/100
Source evidence
56/100
Source signals
70/100
Model compatibility evidence
  • claudeinferred

    The skill text mentions Claude or a closely associated term. This is keyword-derived compatibility; no hands-on model test is recorded.

  • chatgptuntested

    No model-specific signal or recorded compatibility test was found.

  • geminiuntested

    No model-specific signal or recorded compatibility test was found.

  • copilotuntested

    No model-specific signal or recorded compatibility test was found.

  • llamauntested

    No model-specific signal or recorded compatibility test was found.

  • perplexityuntested

    No model-specific signal or recorded compatibility test was found.

  • mistraluntested

    No model-specific signal or recorded compatibility test was found.

  • grokuntested

    No model-specific signal or recorded compatibility test was found.

Inferred compatibility is based on content signals. Check the app, its available tools and setup requirements before running a skill.

What this skill does

# Advisory

Use this skill to manage the full security advisory lifecycle. It enforces atomicity — advisory files, the README index, practices.md, GitHub issues, and the GHSA are updated together so nothing drifts.

Two modes:

- **`/advisory new`** — File a new advisory for a discovered vulnerability - **`/advisory close <slug> <version>`** — Mark an existing open advisory fixed in a release

---

## When to Use

- A vulnerability has been identified and needs a disclosure record - A fix has shipped for an open advisory and the record needs to be closed out - `practices.md` is out of date with the current advisory set

Invoke with `/advisory new` or `/advisory close <slug> <version>`.

---

## Constraints (apply throughout)

- **MUST NOT** include attack vector specifics, exploit steps, or payload details in any repo markdown file (`docs/security/advisories/`, `practices.md`, GitHub issues). Those belong in the GHSA, which is private until coordinated disclosure. - Sanitized content describes: what category of bug, what prerequisites, what data is accessible, what the severity is, and what the fix entails — without explaining *how* to perform the access. - **MUST** commit advisory files, README update, and practices.md update atomically in a single commit. - **MAY** use `SKIP_TESTS=1` on commit *only* when pre-existing, unrelated test failures on the branch would otherwise block the pre-commit hook for this docs-only change. Do not skip tests when your change could plausibly affect them.

---

## Mode: `/advisory new`

### Step 1: Gather inputs

Collect the following before writing anything:

| Field | Notes | |-------|-------| | Slug | `<YYYY-MM-DD>-<short-kebab-description>` — date is discovery date | | CVE class | Category of vulnerability (e.g., tenant isolation bypass, auth middleware gap) | | Affected version(s) | Which released tag(s) contain the bug | | Prerequisites | What the attacker needs — be specific but sanitized | | Impact | What data/functionality is exposed — no exploit steps | | Severity | Low / Medium / High with reasoning | | Remediation approach | What the fix entails — high level | | Tracking issues | GitHub issue numbers for fix and gate gaps (create if they don't exist) | | Gate gap | Which security gate(s) failed to catch this and why |

If any field is unclear, ask before proceeding.

### Step 2: Draft the advisory file

Create `docs/security/advisories/<slug>.md` using this structure:

```markdown # Advisory: <Title>

**Advisory ID:** <slug> **Date:** <YYYY-MM-DD> **Severity:** <Low|Medium|High> (<brief rationale>) **Status:** <Fixed in vX.Y.Z | Open — fix tracked in #N, #M>

---

## CVE Class

<One paragraph: vulnerability category, what makes it exploitable, regression class if known.>

---

## Affected Versions

- vX.Y.Z

<Fixed in / Not yet fixed.>

---

## Prerequisites

<Numbered list of what an attacker needs. Sanitized — no exploit steps.>

---

## Impact

<What data or functionality is exposed. Sanitized — no exploit steps.>

---

## Severity

**<Low|Medium|High>** <reasoning paragraph>

<Escalation condition if applicable.>

---

## Remediation

<Fix approach. Reference tracking issues. No exploit details.>

---

## Disclosure Timeline

| Date | Event | |------|-------| | <date> | Vulnerability identified | | <date> | Advisory filed | | TBD | Fix released |

---

## Tracking Issues

- **#N** — <description>

---

## Gate Gap

<Which gate(s) did not catch this. Why. What the extension entails. Reference tracking issue for gate fix.> ```

### Step 3: Update the advisories README index

Add a row to the index table in `docs/security/advisories/README.md`:

```markdown | [<slug>](./<slug>.md) | <Short title> | vX.Y.Z | <Severity> (<rationale>) | <Open — fix tracked in #N | Fixed in vX.Y.Z> | ```

### Step 4: Update practices.md

Open `docs/security/practices.md` and make the following updates:

1. **For each gate the advisory reveals a gap in:** add a row to that gate's "Known gaps" section describing the gap and referencing the advisory slug.

2. **For each gate extension the advisory motivates:** add a row to that gate's "History" table with the date, change description, and advisory slug.

3. **In the "Advisory → hardening linkage" section:** add a bullet for the new advisory linking it to the specific gate work it motivated (or a waiver if no gate change is needed).

4. **If the advisory introduces an escalation condition** (e.g., single-tenant assumption): update the "Cross-cutting practices" section to document it.

The practices.md update must reflect the *current* state — pending gate work is noted as "Pending, tracked in #N", not omitted.

### Step 5: File sanitized GitHub issues

For each required fix and gate gap that doesn't already have a tracking issue:

```bash gh issue create \ --title "<sanitized title>" \ --body "<sanitized description — no attack vector details>" \ --label "security" ```

Issues describe *what* to fix (add per-user scoping, extend gate coverage) without explaining *how the vulnerability works*.

### Step 6: Commit atomically

Stage all changed files and commit:

```bash git add docs/security/advisories/<slug>.md \ docs/security/advisories/README.md \ docs/security/practices.md

SKIP_TESTS=1 git commit -m "docs(security): file advisory <slug>

<one-line summary of what the vulnerability is>

Advisory covers: <CVE class>. Affected: <version>. Severity: <level>. Fix tracked in #N. Gate gap tracked in #M.

Co-Authored-By: Claude Sonnet 4.6 <[email protected]>" ```

### Step 7: Open a GitHub Security Advisory (GHSA)

This is where full technical details can go — the GHSA is private until coordinated disclosure.

```bash gh api repos/{owner}/{repo}/security-advisories --method POST \ --field summary="<title>" \ --field description="<full technical description including prerequisites and exploit details>" \ --field severity="<low|medium|high|critical>" \ --field vulnerabilities='[{"package":{"ecosystem":"npm","name":"neotoma"},"vulnerable_version_range":"= X.Y.Z"}]' ```

If the API call fails due to token scope, instruct the user to open manually at: `https://github.com/<owner>/<repo>/security/advisories/new`

and provide the full content to paste.

Record the GHSA URL in the advisory file's Disclosure Timeline once it's created.

### Step 8: Push and open a PR

```bash git push origin <branch> gh pr create \ --title "docs(security): file advisory <slug>" \ --base main \ --body "Files sanitized advisory <slug> (category: <vuln class>, affected: <versions>, severity: <level>). No attack-vector specifics, exploit steps, or payloads — those live in the private GHSA. See docs/security/advisories/<slug>.md." ```

The PR description must not expose attack vector details either — treat it as public.

---

## Mode: `/advisory close <slug> <version>`

Use when a fix has shipped in a release and the advisory record needs to be closed.

### Step 1: Verify the fix is released

```bash gh release view <version> ```

Confirm the release tag exists and the fix commits are included.

### Step 2: Update the advisory file

In `docs/security/advisories/<slug>.md`:

- Change `**Status:** Open — fix tracked in #N` to `**Status:** Fixed in <version>` - Fill in the Remediation section with what was actually done (commits, approach) - Add the fix release date to the Disclosure Timeline table

### Step 3: Update the README index

Change the Status cell for the advisory row from `Open — fix tracked in #N` to `Fixed in <version>`.

### Step 4: Update practices.md

For each gate extension that was implemented as part of the fix:

1. Move pending items from "Known gaps" to the "History" table with the date they landed 2. Update the "Advisory → hardening linkage" bullet to show the gate work as completed

### Step 5: Close tracking issues

```bash gh issue close <fix-issue-number> --reason completed gh issue close <gate-gap-issue-number> --reason completed ```

Add a closure comment on each:

```bash gh issue comment <number> --body "Fixed in <version>. Advisory: docs/security/advisories/<slug>.md" ```

### Step 6: Update the GHSA

If the GHSA is open, update it to reflect the fix and set it to published if disclosure is appropriate:

```bash gh api repos/{owner}/{repo}/security-advisories/<ghsa-id> --method PATCH \ --field state="published" ```

### Step 7: Commit and push

```bash git add docs/security/advisories/<slug>.md \ docs/security/advisories/README.md \ docs/security/practices.md

SKIP_TESTS=1 git commit -m "docs(security): close advisory <slug> — fixed in <version>

Co-Authored-By: Claude Sonnet 4.6 <[email protected]>"

git push origin <branch> gh pr create --title "docs(security): close advisory <slug>" --base main ```

---

## File and gate reference

| File | Purpose | Updated by this skill | |------|---------|----------------------| | `docs/security/advisories/README.md` | Index of all advisories | Every invocation | | `docs/security/advisories/<slug>.md` | Per-vulnerability record | `new` and `close` | | `docs/security/practices.md` | Living gate posture and hardening history | Every invocation | | `scripts/security/protected_routes_manifest.json` | Route auth manifest | Only if fix touches manifest | | `tests/security/auth_topology_matrix.test.ts` | Auth-topology test | Only if gate gap fix lands |

---

## Checklist before committing

- [ ] Advisory file has no attack vector details (prerequisites and impact are sanitized) - [ ] README index row added/updated - [ ] practices.md updated: gate gaps noted, history rows added, hardening linkage bullet present - [ ] GitHub issues filed for fix and gate gap (if not pre-existing) - [ ] All three docs staged in the same commit - [ ] GHSA opened or queued for manual creation - [ ] PR description is sanitized (public-safe)

Intended uses

  • A vulnerability has been identified and needs a disclosure record
  • A fix has shipped for an open advisory and the record needs to be closed out
  • practices.md is out of date with the current advisory set

Implementation notes

  • Review the source instructions and adapt inputs before running the workflow.

Try it, then tell us how it went

A copied prompt does not tell us whether the task worked. Report your result after checking the output. We collect your tool and outcome, never your files or prompt.