power
Skill2 starsChanged 31 days agoMITView on GitHub
Initialize Aesop into ANY repo (first-run) or prime the orchestration brain (already-primed). Loads rules, memory, and project state; writes CLAUDE.md layer for unprimed repos.
- Commits and pushes
3 other projects use a version of this.
Adopted by
Most recently taken up by
- hemal08ce094/DesignSkillsSwiftUItheir own version, first seen Nov 8, 2026
- lachinovedhem/standards-mcptheir own version, first seen Oct 11, 2026
- bergside/awesome-design-skillstheir own version, first seen Apr 24, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- bergside/awesome-design-skills★ 2.9ksame name
- hemal08ce094/DesignSkillsSwiftUI★ 0same name
- lachinovedhem/standards-mcp★ 0same 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 repository21 other files it gives its agents.
- auth/CLAUDE.mdCLAUDE.md
- bin/CLAUDE.mdCLAUDE.md
- chat/CLAUDE.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- daemons/CLAUDE.mdCLAUDE.md
- dash/CLAUDE.mdCLAUDE.md
- driver/CLAUDE.mdCLAUDE.md
- hooks/CLAUDE.mdCLAUDE.md
- mcp/CLAUDE.mdCLAUDE.md
- monitor/CLAUDE.mdCLAUDE.md
- scan/CLAUDE.mdCLAUDE.md
- skills/CLAUDE.mdCLAUDE.md
- state_store/CLAUDE.mdCLAUDE.md
- stdio/CLAUDE.mdCLAUDE.md
- tests/CLAUDE.mdCLAUDE.md
- tools/CLAUDE.mdCLAUDE.md
- ui/CLAUDE.mdCLAUDE.md
- buildsystemSkill
- dashboardSkill
- fleetSkill
- healthcheckSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
26 rules. 0 are common, 12 are distinctive, and 14 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
12
Distinctive
This author's own choices. Worth a look.
14
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Core principle: All orchestration rules live on disk as readable .md files.
- Very specific: line 15
Prerequisites: Before running /power, the Aesop harness must be installed in the repository.
- Distinctive: line 28
If the target repository has no CLAUDE.md layer (i.e., no project-root CLAUDE.md, no domain CLAUDE.md units):
- Very specific: line 30
Run /power in the repository root (after CLI setup above)
- Distinctive: line 33
Write a project-root CLAUDE.md (purpose, build/run commands, gotchas)
- Distinctive: line 34
Write smallest self-contained domain CLAUDE.md units (module layout, key contracts, invariants)
- Very specific: line 35
Write a STATE.md checkpoint seed if warranted
- Very specific: line 36
Commit + push the new .md layer (gated by secret-scan)
- Very specific: line 37
Every future /power run then takes the fast path (already-primed).
- Distinctive: line 40
If the repository already has CLAUDE.md layer (project-root + domain CLAUDE.md files ± STATE.md):
- Very specific: line 42
Run /power in the repository root
- Very specific: line 55
Do any domain-scoped CLAUDE.md files exist (e.g., src/CLAUDE.md, api/CLAUDE.md)?
- Distinctive: line 69
Read these directly (they are small and authoritative):
- Very specific: line 84
Read ~/.claude/MEMORY.md (index) and cross-reference any entries relevant to the active project or session task
- Very specific: line 89
Look for STATE.md or BUILDLOG.md in the project root
- Very specific: line 97
Health line first: run python <project-root>/tools/power_selftest.py and lead with its POWER-SELFTEST: line (hooks, brain push state, …
- Distinctive: line 106
After reporting primed, bring the active project's runnable app up in the BACKGROUND if it has one.
- Distinctive: line 111
If healthy (200), skip startup and just report the URL
- Very specific: line 112
If not healthy, start the service in the background via Bash with run_in_background: true
- Distinctive: line 119
If your project uses an orchestration monitor, spin up a background Haiku agent that:
- Distinctive: line 127
If the session is actively developing (writing/changing code — not just answering questions), keep these background loops running …
- Very specific: line 204
Create a STATE.md in the repository root.
- Distinctive: line 230
Commit with a clear message:
- Distinctive: line 240
Run the secret-scan gate:
- Distinctive: line 256
After a successful push, report:
- Very specific: line 276
Portability: all file paths are relative to the target repository or configurable via environment variables / aesop.config.json (never …
“Common” is measured: kept by at least 6 in 10 of the 3 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
Power — Initialize or Prime the Orchestration Brain
This skill makes Aesop work in ANY repository — either initializing a new codebase for the first time (init-prime) or loading an already-primed project's rules and state.
Core principle: All orchestration rules live on disk as readable .md files. This skill reads them, internalizes them,
and reports a compact "primed" summary — enabling autonomous dispatch under the loaded rules.
Prerequisites: Before running /power, the Aesop harness must be installed in the repository. Use the CLI to scaffold:
# Scaffold the Aesop harness (daemons, skills, tools, hooks) into your repo
npx @matt82198/aesop . --name "my-project"
# or run `npx @matt82198/aesop --help` for more optionsThe CLI automatically installs git hooks, copies all necessary directories, and generates initial config files. Then run /power to prime the orchestration brain.
Quick Start: Two Paths
Path 1: First Time? (Init-Prime)
If the target repository has no CLAUDE.md layer (i.e., no project-root CLAUDE.md, no domain CLAUDE.md units):
- Run
/powerin the repository root (after CLI setup above) - Aesop will automatically:
- Fan out read-only Haiku explorers to scan the codebase (briefs only, never raw dumps)
- Write a project-root
CLAUDE.md(purpose, build/run commands, gotchas) - Write smallest self-contained domain
CLAUDE.mdunits (module layout, key contracts, invariants) - Write a
STATE.mdcheckpoint seed if warranted - Commit + push the new
.mdlayer (gated by secret-scan)
- Every future
/powerrun then takes the fast path (already-primed).
Path 2: Already Primed?
If the repository already has CLAUDE.md layer (project-root + domain CLAUDE.md files ± STATE.md):
- Run
/powerin the repository root - Aesop will load the persisted semantic memory from disk (no re-investigation needed)
- Report a compact primed brief with health checks and next steps
- Proceed under the loaded dispatch model
Procedure (in order)
1. Detect priming status
Check if this codebase has been primed before:
- Does a project-root
CLAUDE.mdexist in the target repository root? - Do any domain-scoped
CLAUDE.mdfiles exist (e.g.,src/CLAUDE.md,api/CLAUDE.md)? - Does
STATE.mdor similar checkpoint exist?
If YES → This codebase is already primed. Jump to Step 2: Already-Primed Fast Path.
If NO → This codebase is unprimed. Jump to Step 2: Init-Prime Path.
2A. Already-Primed Fast Path
For an already-primed codebase:
2A.1 Load global rules + dispatch model
Read these directly (they are small and authoritative):
~/.claude/CLAUDE.md— the cardinal rules (autonomous continuation, dispatch model, durable checkpointing, build env)- Project-root
CLAUDE.mdin the current working directory
Then run the brain integrity check: if you version ~/.claude as a git repo, check:
git -C ~/.claude status --porcelainand ahead/behind vs origin. If there are:
- Deliberate uncommitted rule/hook/skill/memory changes → commit + push now (gated by
python <project-root>/tools/secret_scan.py --staged) - Unexpected modifications or deletions → report as alert, stop, diff against remote, restore before continuing
Internalize the dispatch model — document its key invariants locally in your session context (do not modify .md files during this step).
2A.2 Load memory
- Read
~/.claude/MEMORY.md(index) and cross-reference any entries relevant to the active project or session task - For each recalled memory entry, read its full file
- Verify any file/flag/path they name still exists before relying on it
2A.3 Load the active project's state
- Look for
STATE.mdorBUILDLOG.mdin the project root STATE.mdis the source of truth for intent, locked decisions, contracts, and NEXT STEPS- Re-sync from disk (
git log, file state) — never trust stale assumptions
2A.4 Report "primed" (compact)
Output a short brief, not a wall of text:
- Drain the UI inbox first (if configured in
aesop.config.json): runpython <project-root>/tools/inbox_drain.py pendingand surface each item as a one-liner under a "QUEUED FROM DASHBOARD" heading - Health line first: run
python <project-root>/tools/power_selftest.pyand lead with itsPOWER-SELFTEST:line (hooks, brain push state, heartbeats, decisions, scanner state). Bullet anything non-OK. - Open decisions: surface each ACTIVE line from any decisions/proposals file and unactioned inbox entries as one-liners
- Rules loaded: dispatch model + any project-specific overrides in one or two lines
- Project: name, purpose, status, repo, key paths
- Semantics: the domain model + main flow in a few bullets
- Gotchas: the verified ones that change how you act
- Next steps: from
STATE.md, ready to execute (autonomously, per the dispatch model)
2A.5 Project app launch (if applicable)
After reporting primed, bring the active project's runnable app up in the BACKGROUND if it has one. Make it idempotent: if the port is already serving a healthy instance, skip the start and just report the URL.
Pattern (documented in your project's CLAUDE.md):
- Detect if the service is already healthy (e.g.,
curl -s -o /dev/null -w '%{http_code}' http://localhost:PORT/) - If healthy (200), skip startup and just report the URL
- If not healthy, start the service in the background via Bash with
run_in_background: true - Poll readiness WITHOUT blocking sleep: use curl retries or a cheap probe loop
- Report the URL + background task ID so it can be stopped on request
DO NOT hardcode paths. Instead, document in the project's CLAUDE.md how to launch the app and reference that in this skill or a project-specific override.
2A.6 Machinery health (optional)
If your project uses an orchestration monitor, spin up a background Haiku agent that:
- Collects health signals (cheap, deterministic)
- Reads the brief and acts per a charter — AUTO-applies safe refinements and stages rule/behavior changes to
PROPOSALS.mdfor review - Self-paces via ScheduleWakeup and beats a heartbeat
Skip if .monitor-heartbeat is already <300s old. Announce its task id so it can be paused/stopped.
2A.7 Standing dev loops (optional; only during active development)
If the session is actively developing (writing/changing code — not just answering questions), keep these background loops running continuously. They are the "tail": cheap, persistent, append-only.
Patterns (documented in your project's CLAUDE.md):
- Build/status + backup watchdog (background bash loop) — appends timestamped git/test/build snapshots. For example, if your project has a
daemons/folder with arun-watchdog.sh, launch that via Bash withrun_in_background: trueand let it beat a heartbeat - Memory keeper (Haiku, background) — as new decisions, contracts, or gotchas emerge, updates relevant memory files compactly
- Continuous QA loop — after each feature reaches a checkpoint: Haiku reviews → Haiku bugfixes → Haiku lint/format, looping until build is green and review is clean, with Opus orchestrator final-catch before merge
- Tail-drift alignment (Haiku, background) — periodically checks in-flight agents' output against their domain
CLAUDE.mdcontract, realigns if drifting - Hung-Haiku watchdog (orchestrator's main thread, NOT a subagent) — keep watching every spawned Haiku for stalls; on hang: TaskStop it and relaunch from checkpoint
Launch any that aren't already running; stop them with TaskStop when development pauses.
2B. Init-Prime Path
For an unprimed codebase, run the following steps in order:
2B.1 Scan the target repository
Fan out read-only Haiku subagents to explore the codebase structure:
- Haiku A (30s): Repository structure, build system, entry points, main languages
- Haiku B (30s): Key libraries, frameworks, external dependencies
- Haiku C (30s): Testing strategy, CI/CD configuration, deployment targets
- Haiku D (30s): Known gotchas, non-obvious invariants, anti-patterns in the codebase
Each Haiku returns a 1-page brief (never raw code dumps). Collect all four briefs.
2B.2 Synthesize project-root CLAUDE.md
Using the four exploration briefs, write a project-root CLAUDE.md in the target repository.
Include:
- What: 1-line project purpose + primary domain/stack
- Key commands: build, test, run, deploy (exactly as documented in the project)
- Gotchas: verified non-obvious facts that affect how you work
- "Tests must run in isolation (no shared DB state)"
- "Deploy requires manual approval in Slack"
- "Docker build takes 5 minutes; cache aggressively"
- Domain map: 2-4 self-contained scopes, each mapping to one potential Haiku (keep it minimal; don't over-partition)
- See Also: pointers to ARCHITECTURE.md, CONTRIBUTING.md, etc. if they exist in the repo
Keep the project-root CLAUDE.md small (under 50 lines). It's a navigation map, not a textbook.
2B.3 Synthesize domain CLAUDE.md units
For each domain identified in 2B.2, write a scoped domain CLAUDE.md in that module's root directory.
Example structure:
my-repo/
CLAUDE.md <- project-root (what, commands, gotchas, domain map)
src/
CLAUDE.md <- domain 1 (API layer: contracts, invariants, entry points)
api/
...
core/
CLAUDE.md <- domain 2 (core logic: key abstractions, state machine)
...
tests/
CLAUDE.md <- domain 3 (test infrastructure: harness, fixtures)
...Each domain CLAUDE.md includes:
- Module layout: directory tree, what lives where
- Key contracts: public interfaces, APIs, data model
- Domain logic: state machine, algorithm, non-obvious invariants
- Dependencies: external libraries, internal cross-domain calls
- Anti-patterns: what NOT to do in this domain
- Testing: how to run tests, fixtures, mocks
Keep each domain CLAUDE.md focused and minimal (10-20 lines; longer ones should split further).
2B.4 Write STATE.md checkpoint seed
Create a STATE.md in the repository root. This is the durable checkpoint that survives wipes and outages:
# {{PROJECT_NAME}} — State & Next Steps
**Phase**: init-prime (first CLAUDE.md layer written)
**Intent**: {{PROJECT_NAME}} is now initialized with aesop orchestration semantics.
All domain rules live in scoped CLAUDE.md files. Ready to dispatch first work.
**Locked decisions**:
- (none yet; updated as decisions are made)
**NEXT STEPS**:
- [ ] Review + approve the written CLAUDE.md layer
- [ ] Run `/buildsystem` to execute one example orchestration wave
- [ ] Iterate domain CLAUDE.md files as you discover invariants2B.5 Commit + Push
Stage all newly written files:
git add CLAUDE.md STATE.md src/CLAUDE.md core/CLAUDE.md tests/CLAUDE.mdCommit with a clear message:
git commit -m "init: initialize aesop orchestration semantics — project CLAUDE.md + domain units + STATE.md seed
Aesop init-prime: generated project-root CLAUDE.md (purpose, commands, gotchas),
minimal domain CLAUDE.md units (contracts, invariants), and STATE.md checkpoint seed.
Co-Authored-By: Aesop Orchestrator <aesop@example.com>"Run the secret-scan gate:
python <project-root>/tools/secret_scan.py --stagedIf gate fails → fix the leaky files and re-commit (never override the gate).
If gate passes → push to the feature branch:
git push origin feat/aesop-init-prime(Or to main if you have override permissions, but best practice: feature branch → PR → review → merge.)
2B.6 Report init completion
After a successful push, report:
✓ Aesop init-prime complete
- Project-root CLAUDE.md written (60 lines; purpose, commands, gotchas, 3 domains)
- Domain CLAUDE.md units: src/, core/, tests/
- STATE.md checkpoint seed created
- Pushed to origin feat/aesop-init-prime (gated by secret-scan, all clear)
NEXT: Review the generated `.md` layer, iterate on domains, then run /buildsystem for your first orchestration wave.Notes
- If no project context exists yet (e.g., you're answering a general question about rules): Steps 1 loads global rules; skip 2 and say so.
- Steps 1-5 always run on
/power; step 6 (launch app) runs only if the active project has one; step 6b (monitor) runs only if configured; step 7 (dev loops) runs only during active development. - Init-prime writes
.mdfiles to the target repository (never personal/system paths). All state lives under the target repo root or configuredstate_root(seeaesop.config.json). - Prefer subagents for anything large; the goal is a brain, not a dump.
- After priming (either path), proceed under the loaded dispatch model (continue approved work autonomously; stop only for genuine decisions or unauthorized outward/destructive actions).
- Portability: all file paths are relative to the target repository or configurable via environment variables /
aesop.config.json(never hardcoded personal paths). Use~notation for home-directory paths (e.g.,~/.claude,~/scripts) for cross-platform compatibility (Windows, macOS, Linux). Config loaders automatically expand~at runtime.
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.

