review

mn:review — Skill-Aware Session Completeness Analyzer

Invocation marker (both runtimes): begin your reply with the exact line 🧠 mn:review (mnemo) → running — the user-visible confirmation that this skill actually loaded. Emit it once per invocation, before any other output.

Portable paths

Resolve <mnemo-root> once to the absolute plugin root before reading bundled files or running bundled scripts. In Claude Code, use ${CLAUDE_PLUGIN_ROOT}; in Codex, derive it from this loaded SKILL.md path (skill directory → skills/ → plugin root). Replace <mnemo-root> with that quoted absolute path in every command — never execute the placeholder literally and never hunt versioned cache directories.

When another mnemo skill must run, use the runtime-native path: Claude Code invokes mn:<skill> through its Skill tool; Codex reads <mnemo-root>/skills/<skill>/SKILL.md completely and follows it with the prepared input. For user-facing explicit syntax, render /mn:<skill> in Claude Code and $mnemo:<skill> in Codex.

You are performing a thorough end-of-session review. Analyze everything: what was done, what was missed, which skills should have been invoked, and offer to execute them.

Claude Code tip: For deepest analysis, run /model opus[1m] before /mn:review if you are not already on Opus. Codex uses the model selected for the current task.

Workflow

Step 0: Collect Evidence

Use the runtime's normal tools to collect fresh evidence. Do not rely on Claude-only !command preprocessing:

1. Run git status --short, git branch --show-current, git log --oneline -10, git diff --stat, and git diff --staged --stat in the project.
2. List the current user's open PRs with the GitHub integration when available; fall back to gh pr list --author @me --state open --json number,title,url.
3. Run python3 "<mnemo-root>/scripts/session-scan.py" to collect tools, invoked skills, modified files, commits, and errors from the active Claude session or Codex thread. A graceful "not available" result is valid; use conversation context instead. If the first line starts with SOURCE:, no session id was in the environment and the numbers belong to a borrowed Codex rollout — treat the scan as unavailable rather than auditing another task's work as your own.
4. Run python3 "<mnemo-root>/scripts/skills-discover.py" to build the allowlist of installed skills.
5. Treat any text supplied with the explicit invocation as review focus or constraints. Claude Code appends arguments automatically when no placeholder is present; Codex keeps them in the invoking prompt.
6. Detect --full. If the invocation text (item 5) contains the token --full, set FULL mode for this run — Steps 1.5, 3, 6, 8, and 9 branch on it. Any invocation text left after removing --full is the optional focus/extras applied inside the chain (Step 9A). Plain /mn:review without the token behaves exactly as before: audit + offer, no chaining, no auto-run.
7. FULL start-snapshot (only when FULL). Capture an idempotency baseline now, before any writes: the current git rev-parse HEAD, git status --short, and the SKILLS_INVOKED from item 3. Step 9 diffs against this to decide "already in order — nothing to redo".