cold-start
Skill3 starsChanged 4 months agoApache-2.0View on GitHub
Bootstrap the ai/ knowledge layer for a repo that has none — draft the MODULE_MAP, diagrams, and guide docs as [inferred] for a human to audit, touching no source code. Use when the maps are still placeholders or the user asks to cold-start, bootstrap, or onboard a repository.
6 other projects use a version of this.
Adopted by
Most recently taken up by
- ZwaarContrast/skillstheir own version, first seen Nov 1, 2026
- garrytan/gbraintheir own version, first seen Oct 17, 2026
- Pearl1531/Archetype-talktheir own version, first seen Oct 13, 2026
- serithemage/serverless-openclawtheir own version, first seen Oct 8, 2026
- yogsoth-ai/north-star-crystallizationtheir own version, first seen May 16, 2026
- vivy-yi/xiaohongshu-skillstheir own version, first seen Jan 22, 2026
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- garrytan/gbrain★ 33ksame name
- vivy-yi/xiaohongshu-skills★ 455same name
- serithemage/serverless-openclaw★ 196same name
- Pearl1531/Archetype-talk★ 2same name
- yogsoth-ai/north-star-crystallization★ 2same name
- ZwaarContrast/skills★ 1same name
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 repository32 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- .cursor/rules/add-feature.mdcCursor rule
- .cursor/rules/adversarial-audit.mdcCursor rule
- .cursor/rules/ai-knowledge-layer.mdcCursor rule
- .cursor/rules/check-drift.mdcCursor rule
- .cursor/rules/cold-start.mdcCursor rule
- .cursor/rules/create-feature-catalog.mdcCursor rule
- .cursor/rules/fix-bug.mdcCursor rule
- .cursor/rules/perform-feature-add-simulation.mdcCursor rule
- .cursor/rules/post-cold-start-verification.mdcCursor rule
- .cursor/rules/review-agent-config.mdcCursor rule
- .cursor/rules/review-change.mdcCursor rule
- .cursor/rules/verify-ai-readiness.mdcCursor rule
- add-featureSkill
- deep-testSkill
- fix-bugSkill
- review-changeSkill
- add-featureSkill
- adversarial-auditSkill
- check-docsSkill
- check-driftSkill
- cold-startSkill
- create-feature-catalogSkill
- fix-bugSkill
- implement-specSkill
- perform-feature-add-simulationSkill
- post-cold-start-verificationSkill
- review-agent-configSkill
- review-changeSkill
- verify-ai-readinessSkill
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 0 are distinctive, and 5 are very specific to this project.
0
Common
Most of the 5 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
Read ai/repo-profile.json (deterministic output of the kit's orient step).
- Very specific: line 24
If repo-profile.json → maturity.process is 2, CLAUDE_bkp_*.md / AGENTS_bkp_*.md exist at the repo root with the team's previous AI …
- Very specific: line 26
Read them, extract KNOWLEDGE — not raw text — (build commands, conventions, architecture notes, gotchas, module descriptions) and merge it …
- Very specific: line 46
If the build/test commands you verified in config files differ from the Build/Test lines stamped into CLAUDE.md and AGENTS.md, correct …
- Very specific: line 68
Do NOT modify any source file.
“Common” is measured: kept by at least 6 in 10 of the 5 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
<!-- Copyright (c) 2026 Kunal Suri (CEA LIST). All rights reserved. -->Cold-start the knowledge layer
The contract: read-and-write-docs-only — touch no source code, tag everything
[inferred], and stop at an AUDIT TODO for the human. You draft the map; a human
verifies it before any feature is built.
1. Load the facts
Read ai/repo-profile.json (deterministic output of the kit's orient step). Treat
its stack facts as given; VERIFY its build/test commands against the real config
files before writing them anywhere as confirmed. If the profile says this repo is a
fork, separating OUR code from FROZEN upstream code is the single most important
output of this pass. If a humanContext block is present, use it:
developer.skillLevel/codebaseFamiliarity calibrate how conservative your
Stability guesses are; stack.kind tells you whether to map a frontend/backend
split; record branch/firstRunAt in PROJECT_OVERVIEW.md.
2. Absorb prior config (Process 2 repos only)
If repo-profile.json → maturity.process is 2, CLAUDE_bkp_*.md /
AGENTS_bkp_*.md exist at the repo root with the team's previous AI configuration.
Read them, extract KNOWLEDGE — not raw text — (build commands, conventions,
architecture notes, gotchas, module descriptions) and merge it into
ai/guide/CONVENTIONS.md, ai/guide/ARCHITECTURE.md, ai/guide/MODULE_MAP.md
(seed rows), and ai/guide/PROJECT_OVERVIEW.md. Tag every extract
[inferred — from prior config].
3. Explore cheaply, read-only
List the tree 2 levels deep, read build manifests (not source), scan the last ~30 commit subjects for active areas. Prefer grep + line counts over whole-file reads. Modify nothing in this step.
4. Fill MODULE_MAP
ai/guide/MODULE_MAP.md: one row per module/package — directory, a one-line
responsibility, the entry-point file, and a Stability GUESS. Use the ACTUAL names
you found; do not assume names. Each row has FIVE cells — Directory,
Responsibility, Entry point, Stability guess, Status — and the Status cell is
where the [inferred] tag goes (the tooling reads provenance from the LAST
cell only).
5. Correct the stamped commands
If the build/test commands you verified in config files differ from the
Build/Test lines stamped into CLAUDE.md and AGENTS.md, correct exactly
those lines (and the Definition-of-done line in ai/guide/CONVENTIONS.md),
tag the correction [inferred], and change nothing else in those files. This
is the sanctioned exception to the No-Churn rule.
6. Draft diagrams
Into ai/analysis/diagrams/ as Mermaid .mmd: package-deps.mmd (dependency
graph), domain-core.mmd (core types), seam.mmd (the main boundary — note the
protocol as a question if unverified).
7. Note features, update the guides
Note candidate features in ai/guide/FEATURE_MAP.md using its template. Update
ai/guide/ARCHITECTURE.md and ai/guide/PROJECT_OVERVIEW.md only where you
VERIFIED something in code or config.
8. Hard rules for this pass
- Tag EVERYTHING you write
[inferred]. You are guessing; say so. - Separate OBSERVED facts (file A imports B) from INFERENCES (A is "the domain layer").
- On a fork: mark anything inherited as Stability
frozenand FLAG it "UNSURE — needs human" rather than asserting it. - Where you cannot tell, write
?and "UNSURE — needs human". Never guess confidently. - Do NOT modify any source file. Only write inside
ai/. - Re-run safety: leave rows already carrying
[verified]unchanged; only populate rows still?or holding placeholder text like<fill in>.
9. Stop and report
Print an "AUDIT TODO" table (columns: # · Location · What to verify · Why uncertain)
covering every row still ?, every frozen guess needing confirmation, and any
protocol/command assumption still [inferred]. The human sets Stability and flips
[inferred] → [verified]. Do NOT proceed to building features. Tell the user they
can run /review-agent-config (structural check of CLAUDE.md/AGENTS.md) and
/post-cold-start-verification (deep semantic audit) once their audit is done.
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.

