agentleFS
grace-methodgrace-method/grace-plugin/skills/install/SKILL.md

install

Skill0 starsChanged 2 months agoMITView on GitHub

Write GRACE methodology rules into the current project. Copies GRACE.md and the six rule files into .claude/, ensures the anchor line is present in CLAUDE.md, and records the installed methodology version. Safe to re-run after a plugin update. Use when setting up a new project with GRACE or when refreshing methodology files after an update.

22 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use update and release.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • AnthropicSkill

    update

    Sync tasks and refresh memory from your current activity. Use when pulling new assignments from your project tracker into TASKS.md, triaging stale or overdue tasks, filling memory gaps for unknown people or projects, or running a comprehensive scan to catch todos buried in chat and email.

    Also in 3 projects that use this

  • MicrosoftSkill

    release

    Use ONLY when the user explicitly asks to cut/publish a Huabu desktop release or tag a version. Performs an outward action — bumps the version in apps/desktop/package.json, commits + annotated-tags vX.Y.Z, and pushes the branch + tag. That tag push auto-builds the macOS/Windows installers and publishes a GitHub Release. Provides `status`, `bump`, `tag-push`, and a one-command `run` (bump → tag-push).

    Also in 3 projects that use this

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. 0 are common, 3 are distinctive, and 13 are very specific to this project.

  • 0

    Common

    Most of the 18 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 13

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Follow these steps in order.

    line 22
  2. Very specific:

    If neither has both files, STOP: Write nothing. Report every location you tried and what was missing from each, and tell the user to run …

    line 35
  3. Very specific:

    Create .claude/ and .claude/rules/ if they do not already exist.

    line 53
  4. Very specific:

    Read methodology/GRACE.md from the payload root resolved in Step 0.

    line 57
  5. Very specific:

    Write its contents to .claude/GRACE.md in the current project, overwriting any existing file.

    line 59
  6. Very specific:

    Read each of the six rule files from methodology/rules/ in the payload root:

    line 65
  7. Very specific:

    Write each to .claude/rules/<filename> in the current project, overwriting any existing file.

    line 73
  8. Very specific:

    Check .claude/CLAUDE.md in the current project:

    line 81
  9. Distinctive:

    If the file does not exist: Create it containing only the anchor line.

    line 83
  10. Very specific:

    Never modify any other content in .claude/CLAUDE.md.

    line 92
  11. Very specific:

    Read methodology-freshness.json from the payload root resolved in Step 0.

    line 97
  12. Very specific:

    When it is true, also tell the user plainly that the plugin was built from an uncommitted tree, so the recorded revision names a real …

    line 112
  13. Very specific:

    Write two files into .claude/, describing exactly the seven methodology files this procedure wrote — GRACE.md and the six rules.

    line 120
  14. Very specific:

    Do not touch anything outside .claude/ in the current project.

    line 180
  15. Very specific:

    Do not read or modify .git/, source files, build outputs, environment files, or any content outside .claude/.

    line 181
  16. Distinctive:

    Do not delete any files. This skill only creates or overwrites the specific files listed above.

    line 183

“Common” is measured: kept by at least 6 in 10 of the 18 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Install GRACE Methodology Rules

This skill writes the GRACE methodology files into the current project so they are loaded into every conversation via the @-import chain. It is the counterpart to the plugin's skill distribution: skills travel with the plugin automatically, but rules require on-disk files because plugins cannot ship always-on CLAUDE.md content.

What This Skill Does

Follow these steps in order. Do not skip steps. Report what you did at the end.

Step 0 — Resolve the payload root, or stop

Everything below reads from a payload root: a directory containing both methodology-freshness.json and methodology/. Find it by trying these in order, and take the first that has both:

  1. Two levels up from this SKILL.md. This is the plugin layout — the directory holding the skills/ folder this file sits in.
  2. ~/.grace/. This is the local-runtime layout, deposited by ./install.sh on a machine that installs from source rather than the marketplace.

If neither has both files, STOP. Write nothing. Report every location you tried and what was missing from each, and tell the user to run ./install.sh from the GRACE repository if this is a local-runtime machine.

Do not search for a payload root, and do not use one you were not given by this list. An agent told to "find the plugin root" will search the filesystem, find some checkout of the methodology repository, and write a project from a source nobody verified — and that failure looks exactly like success. A skill that stops can be recovered from; a skill that guesses cannot even be detected. This is defect D-10, and Step 2 below is where it used to fail silently: on a local-runtime machine two levels up resolves to ~/.claude/, which holds only rules/ and skills/, so there was never a methodology/GRACE.md to read.

Refer to the directory you resolved here as the payload root for the rest of this procedure.

Step 1 — Create directories

Create .claude/ and .claude/rules/ if they do not already exist.

Step 2 — Write GRACE.md

Read methodology/GRACE.md from the payload root resolved in Step 0.

Write its contents to .claude/GRACE.md in the current project, overwriting any existing file. This file is owned by the methodology; do not prompt for confirmation before overwriting.

Step 3 — Write rule files

Read each of the six rule files from methodology/rules/ in the payload root:

  • principles.md
  • consultant-reflex.md
  • interaction.md
  • implementation-discipline.md
  • test-discipline.md
  • version-control.md

Write each to .claude/rules/<filename> in the current project, overwriting any existing file. These files are owned by the methodology; do not prompt for confirmation before overwriting.

Step 4 — Ensure anchor line in CLAUDE.md

The anchor line is: @.claude/GRACE.md

Check .claude/CLAUDE.md in the current project:

  • If the file does not exist: Create it containing only the anchor line.
  • If the file exists and already contains the anchor line (a line matching the regex ^@\.claude/GRACE\.md\b): Do nothing to the file.
  • If the file exists and does NOT contain the anchor line: Append the anchor line at the end of the file, preceded by a blank line.
  • If the file contains a near-match (a line containing GRACE.md in an @ import but not matching the exact pattern above): Report the mismatch to the user and ask them to confirm before replacing it.

Never modify any other content in .claude/CLAUDE.md. That file is owned by the user.

Step 5 — Record methodology version

Read methodology-freshness.json from the payload root resolved in Step 0. Write or overwrite .claude/grace-methodology-version in the current project with a short summary:

grace-methodology-version: <version from methodology-freshness.json>
installed: <current ISO timestamp>
revision: <repositoryRevision from methodology-freshness.json>
provenance: <clean | DIRTY — revision does not describe installed content>

The provenance line is required. Read workingTreeDirty from methodology-freshness.json: write clean when it is false, and DIRTY — revision does not describe installed content when it is true.

When it is true, also tell the user plainly that the plugin was built from an uncommitted tree, so the recorded revision names a real commit whose content is not what was installed. Copying provenance known to be false is the defect this line exists to prevent (D-01, requirement 8.9); a marker that looks valid and is wrong is worse than one that admits it cannot be substantiated.

Step 6 — Record what you wrote

Write two files into .claude/, describing exactly the seven methodology files this procedure wrote — GRACE.md and the six rules. Not CLAUDE.md, which is the user's, and not the version marker. Together they are what lets ./grace-runtime-status.sh, run from inside this project later, answer "has this copy been edited since it was installed?" — a question that is otherwise unanswerable here, so a project without them is reported as UNVERIFIABLE rather than as current.

.claude/.grace-installed — one path per line, relative to .claude/, naming what GRACE owns in this project and nothing else:

GRACE.md
rules/principles.md
rules/consultant-reflex.md
rules/interaction.md
rules/implementation-discipline.md
rules/test-discipline.md
rules/version-control.md

.claude/.grace-installed.sha256 — the same seven files with their hashes, in shasum -a 256 output format, generated after the writes above have completed so it describes the bytes actually on disk.

Derive it from the ownership list you just wrote, never from a glob:

cd .claude && tr '\n' '\0' < .grace-installed | xargs -0 shasum -a 256 > .grace-installed.sha256

Do not use shasum -a 256 GRACE.md rules/*.md. That glob matches every .md in rules/, including rules the user wrote themselves — putting a file GRACE never wrote into GRACE's own content record. The moment the user edits their own rule, the status tool reports "deployed content differs from what this install wrote", which is false in every word for that file, and the remedy it offers re-runs this skill and launders their edit into GRACE's record. The machine layer had exactly this bug and stopped using a glob over the shared rules directory for exactly this reason. The invariant is in install.sh: "No entry exists for a file GRACE did not write."

Neither file lists itself, and neither lists CLAUDE.md or grace-methodology-version. The two files therefore have the same number of entries — seven. If the receipt has more, a glob has crept back in. Ownership is per file, never per directory: anything of the user's in .claude/ is absent from both lists and is therefore never touched by any later run.

Step 7 — Report

Print a one-paragraph summary of what was written, including:

  • The number of files written: GRACE.md + 6 rules + the version marker = 8, plus the two ownership records from Step 6, which describe the seven methodology files and not themselves.
  • Whether the anchor line was already present or was added.
  • A reminder that .claude/GRACE.md and .claude/rules/*.md are owned by GRACE and will be overwritten on the next /grace:install run.
  • A reminder that the user's .claude/CLAUDE.md is their own; only the anchor line is managed.

Important constraints

  • Do not touch anything outside .claude/ in the current project.
  • Do not read or modify .git/, source files, build outputs, environment files, or any content outside .claude/.
  • Do not delete any files. This skill only creates or overwrites the specific files listed above.
  • The payload root is whatever Step 0 resolved, and only that. Do not re-derive it later in the procedure, and do not fall back to path traversal if a read fails — traversal is Step 0's first rung and the local runtime is its second. If Step 0 could not resolve a root, the procedure has already stopped.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.