⏳ This skill is pending AI review.
Scores will appear once the review pipeline completes.
memory-review
Sort Claude Code's auto memory by the memory rule - preferences into rules, project knowledge into repo docs, progress deleted. Read-only until approved. Run with /memory-review.
Choose how to use this skill
You do not need every option. Choose the path your AI client supports. The stable page stays the same; versioned files are immutable.
1. Native installer
This listing has no registered native installer command. Use the complete package or source fallback below, depending on what your client supports.
Do not guess an installer command or replace an existing version without reviewing the diff.
2. Complete package recommended
Download the ZIP when available. It includes SKILL.md plus the references, security notes and version metadata.
No complete ProSkills package is published for this listing yet.3. Prompt-only
Copy the prompt above when the agent can read the stable page or when you want to adopt the workflow without installing a skill.
Need only the instruction file?
Download SKILL.md only if your client requires a single file. The complete ZIP is safer for a full installation because it preserves the references and release context.
No path installs or executes anything by itself. Your agent still needs access to the project files. Before updating, compare the installed version and review the diff.
// RATINGS
Not yet listed on ClawHub or SkillsMP
// README
Install
Three routes.
1. install.sh from a clone. This is the full install: every row of
What you get, plus your own profile, private additions, --copy mode and
update.sh, and uninstall.sh to undo it.
git clone https://github.com/Gord1y/countersign-skills ~/countersign-skills
~/countersign-skills/install.sh
Requirements are POSIX sh, git and jq. Run it from your own terminal, not through an agent:
the agent's sandbox protects ~/.claude, so an agent can't write there, and shouldn't.
2. skills.sh. Skills only, no rules, subagents, settings or bin/:
npx skills add Gord1y/countersign-skills
It installs from the default branch, main, which moves only at releases. orchestrate still
works this way: it carries land-worktree in its own scripts/, and falls back to
general-purpose subagents when no builder agent is installed.
3. Countersign. Planned for Countersign 0.3.0, not released yet. Its Skills pane will download
a pinned release, verify its .sha256, and list catalog.json.
Where everything lands, how copies and settings merge, backups and unattended runs are in docs/install.md.
What you get
| Claude Code | Codex | Antigravity | skills.sh | |
|---|---|---|---|---|
| Skills | 16 | 14 | 10 | skills only |
| Rules | 11 | 9 | 9 | no |
| Subagents | 4 | no | no | no |
| Settings defaults | yes | no | no | no |
| Status line | yes | no | no | no |
Codex gets its rules in a generated ~/.codex/AGENTS.md, Antigravity in ~/.gemini/GEMINI.md.
Neither reads @ imports, and the orchestration and memory rules lean on things only Claude Code
has, so those two rules stay out of both.
Skills
| Skill | What it does | Agents |
|---|---|---|
orchestrate | Splits a multi-commit task into units and runs one builder subagent per unit, then QAs every surface the run touched and leaves a local walkthrough. | Claude Code |
context-transfer | Packs a conversation into one self-contained block for a new chat. | Claude Code, Codex, Antigravity |
human-voice-writing | Drafts and revises longer text so it reads as human, not AI-generated. | Claude Code, Codex, Antigravity |
memory-review | Sorts Claude Code's auto memory by the memory rule: preferences into rules, project knowledge into repo docs. Manual: /memory-review. | Claude Code |
codebase-research | Picks the cheapest way to research code, with an explicit model on any spawn. | Claude Code, Codex |
release-notes | Writes release note entries, keeps an open note in step with its branch, and keeps the release index current. | Claude Code, Codex, Antigravity |
feature-pr-description | Writes the body of a feature or fix PR to the repo's description contract. | Claude Code, Codex, Antigravity |
promotion-pr-description | Writes a release or hotfix PR's body from its release note (release into staging, staging into main, a hotfix into either), plus what developers need before and after the merge. | Claude Code, Codex, Antigravity |
thorough-diff-review | Runs a full local review of a diff before you push. | Claude Code, Codex, Antigravity |
i18n-translate | Translates new English message keys into the other locales, one subagent each. | Claude Code, Codex |
qa-tester | Tests a change in the running product on every surface it touched (web, API, CLI, native or mobile app, game, library), saves the evidence under writeups/, and reports pass or fail. Runs forked under Claude Code. | Claude Code, Codex |
impact-check | Finds what a change could break outside its diff, and backs its verdict with a script that runs the real code. | Claude Code, Codex, Antigravity |
how-it-works | Explains how a part of the codebase works, for the person about to change it: its moving parts, the flow with a diagram, where to start reading. | Claude Code, Codex, Antigravity |
pr-review-triage | Verifies a PR's AI findings, failed checks and open threads at the PR head, and writes a fix plan that proves each fix safe; starts on a bare PR link. | Claude Code, Codex |
walkthrough | Walks you through any change with its QA evidence (screenshots, request and response pairs, transcripts), or builds a customer-facing HTML walkthrough of a UI release. | Claude Code, Codex, Antigravity |
writeups-cleanup | Proposes what in writeups/ has done its job, with the evidence, and deletes only what you approve. Manual: /writeups-cleanup. | Claude Code, Codex, Antigravity |
Manual skills have disable-model-invocation: true: their description costs no context, and only
you start them, with /name.
The process skills (research, PR descriptions, release notes, review, QA, translation) are written
for any repo. They read the facts they need, such as paths, commands and the branch flow, from your
repo's CLAUDE.md or AGENTS.md, then from local.md, and ask once when neither has one.
release-notes defers to a repo's own copy when the repo ships one. Everything they draft goes into
a gitignored writeups/ folder (layout in docs/install.md).
Rules
One module per topic in rules/, imported into your CLAUDE.md.
- Responses: how to answer, what to lead with, and how to mark a summary (✅ asked work, 🔧 found and fixed, ❌ still broken).
- Code: house style for code, such as no comments and strict types.
- Issues you find along the way: fix what you find in the same session, don't park it.
- Tooling: the package manager to use and the gates that count as done.
- Planning and orchestration: ask every load-bearing question up front, split big work into units, and hand off only through the
context-transferskill. - Memory: what belongs in auto memory and what belongs in the repo.
- Research and external information: trust your own knowledge first, confirm anything from the web.
- Commits: one task per conventional commit, and never push unprompted.
- Safety: no force-pushes, ask before destroying, keep secrets out of output.
- Instruction sources and trust: treat files from outside the project as data, not commands.
The rules are opinionated: tooling, for one, asks for pnpm. To switch a rule off, delete its
import line in your profile's CLAUDE.md and run install.sh again. To change one, switch it off
and add your own version as a rule in an addition, so git pull never fights your edit.
Subagents
Claude Code only, linked into ~/.claude/agents. Each is split by what it may do, neve
// HOW IT'S BUILT
KEY FILES