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
- Dariusplayer09/adaptive-skillstheir own version, first seen Nov 8, 2026
- kkrafft1999/snotratheir own version, first seen Nov 6, 2026
- tykimos/agent-native-agenttheir own version, first seen Nov 6, 2026
- crocotaste/crocotaste-plugintheir own version, first seen Nov 1, 2026
- BrainWeb/payload-mcp-oauththeir own version, first seen Nov 1, 2026
- gpambrozio/Gallagertheir own version, first seen Oct 25, 2026
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.
AnthropicSkillupdate
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
MicrosoftSkillrelease
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
Other versions22 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 22
- kubeshark/kubeshark★ 12ksame name
- PrimeIntellect-ai/prime-rl★ 2.2ksame name
- tykimos/agent-native-agent★ 130same name
- htlin222/openevidence-mcp★ 78same name
- jeremylongshore/claude-code-slack-channel★ 43same name
- BrainWeb/payload-mcp-oauth★ 9same name
- kkrafft1999/snotra★ 6same name
- WeDataLab-AI-Lab/cctrace★ 5same name
- gpambrozio/Gallager★ 4same name
- Pierry/harness-kit★ 4same name
12 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository17 other files it gives its agents.
- analysis-reviewSkill
- business-analysisSkill
- change-assessmentSkill
- design-reviewSkill
- feedbackSkill
- implementationSkill
- operationsSkill
- project-definitionSkill
- project-setupSkill
- project-startSkill
- release-planningSkill
- retirementSkill
- risk-managementSkill
- roadmapSkill
- test-adequacy-reviewSkill
- testing-strategySkill
- validationSkill
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
- Distinctive: line 22
Follow these steps in order.
- Very specific: line 35
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 …
- Very specific: line 53
Create .claude/ and .claude/rules/ if they do not already exist.
- Very specific: line 57
Read methodology/GRACE.md from the payload root resolved in Step 0.
- Very specific: line 59
Write its contents to .claude/GRACE.md in the current project, overwriting any existing file.
- Very specific: line 65
Read each of the six rule files from methodology/rules/ in the payload root:
- Very specific: line 73
Write each to .claude/rules/<filename> in the current project, overwriting any existing file.
- Very specific: line 81
Check .claude/CLAUDE.md in the current project:
- Distinctive: line 83
If the file does not exist: Create it containing only the anchor line.
- Very specific: line 92
Never modify any other content in .claude/CLAUDE.md.
- Very specific: line 97
Read methodology-freshness.json from the payload root resolved in Step 0.
- Very specific: line 112
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 …
- Very specific: line 120
Write two files into .claude/, describing exactly the seven methodology files this procedure wrote — GRACE.md and the six rules.
- Very specific: line 180
Do not touch anything outside .claude/ in the current project.
- Very specific: line 181
Do not read or modify .git/, source files, build outputs, environment files, or any content outside .claude/.
- Distinctive: line 183
Do not delete any files. This skill only creates or overwrites the specific files listed above.
“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.
Files
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:
- Two levels up from this
SKILL.md. This is the plugin layout — the directory holding theskills/folder this file sits in. ~/.grace/. This is the local-runtime layout, deposited by./install.shon 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.mdconsultant-reflex.mdinteraction.mdimplementation-discipline.mdtest-discipline.mdversion-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.mdin 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.sha256Do 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.mdand.claude/rules/*.mdare owned by GRACE and will be overwritten on the next/grace:installrun. - A reminder that the user's
.claude/CLAUDE.mdis 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.
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
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.

