aevum
CLAUDE.md4 starsChanged 41 days agoApache-2.0View on GitHub
A Claude Code orchestration harness for long-form, unattended multi-agent work. Aevum itself is configuration, shell, and Python; the projects that consume it ship the product code. The enforceable rules are in .claude/rules/ (index: .claude/rules/00-index.md) and load automatically. Read these when the task touches them, not up front: - .claude/scripts/orchestrate-block.prompt.md: the inner orchestrator's operating manual (§0 to §3 for the contract, §6 cluster loop, §8 rotation, §10 turn endings). - .claude/templates/block.yaml: block schema. - docs/design-notes.md: why the harness is shaped this…
The file
We split this CLAUDE.md into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 62
Every block pairs docs/blocks/<ID>/block.yaml with a narrative plan under docs/plans/.
- Very specific: line 64
When a change adds, removes, or renames a shipped harness file, update OBSOLETE or the copied directories in scripts/aevum-install.py, the …
- Distinctive: line 69
Never push, force-push, or bypass hooks.
≈ “Common” is an estimate: at least 1 in 5 of the 28,029 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Aevum
A Claude Code orchestration harness for long-form, unattended multi-agent work. Aevum itself is configuration, shell, and Python; the projects that consume it ship the product code.
The enforceable rules are in .claude/rules/ (index:
.claude/rules/00-index.md) and load automatically. Read these when
the task touches them, not up front:
.claude/scripts/orchestrate-block.prompt.md: the inner orchestrator's operating manual (§0 to §3 for the contract, §6 cluster loop, §8 rotation, §10 turn endings)..claude/templates/block.yaml: block schema.docs/design-notes.md: why the harness is shaped this way, and how each choice maps to Anthropic's guidance.docs/swap-points.md: the stack-bound seams.
Layout
.claude/: committed harness config only (rules, hooks, scripts, templates, agents, skills, settings, state.yaml). Runtime writes here are rejected byclaude-dir-write.sh.logs/: runtime artefacts (ledgers, gate verdicts, locks). Gitignored.docs/blocks/<BLOCK_ID>/: block plans and baselines.docs/plans/: narrative plans with YAML frontmatter; closed plans move toarchive/plans/in the same commit..claude-plugin/andskills/init/: the plugin packaging./aevum:initrunsscripts/aevum-install.py, which vendors the core into a project. Plugins cannot ship deny lists or rules, so the core stays project-local by design.
Commands
python3 scripts/block-preflight.py <BLOCK_ID> # validate a block
bash scripts/new-block.sh <BLOCK_ID> "<Title>" <owner> # block.yaml + paired plan
bash .claude/scripts/orchestrate-block-loop.sh <BLOCK_ID> [--dry-run]
bash .claude/scripts/orchestrate-block.sh <BLOCK_ID> --self-test
python3 scripts/aevum-install.py <TARGET_DIR> # dry run; --apply to write
claude plugin validate . # plugin + marketplace manifestsSmoke test for the EXAMPLE block: set its base_sha to a real commit
first; a healthy result is exactly one blocker, status is 'draft'.
Agents and skills
Five infrastructure agents in .claude/agents/, all dispatched by the
block orchestrator: code-reviewer (Gate 3b), config-validator
(Gate 2), criteria-checker (Gate 3a), fix-bucketer (fix loop),
status-tracker (block close). Project specialists are authored from
examples/agents/AGENT_TEMPLATE.md.
User-invoked skills in .claude/skills/: /new-block, /dispatch,
/signoff (namespaced /aevum:<name> when installed as a plugin).
Conventions
- English for code, commits, docs. No em-dashes in anything written for the repo.
- Every block pairs
docs/blocks/<ID>/block.yamlwith a narrative plan underdocs/plans/; scaffold both withscripts/new-block.sh. - When a change adds, removes, or renames a shipped harness file,
update
OBSOLETEor the copied directories inscripts/aevum-install.py, the version in.claude-plugin/plugin.json, andCHANGELOG.md. - Discuss before adding packages.
- Never push, force-push, or bypass hooks; the deny list enforces it.
Promoting an integration branch to
mainis a human action.
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.

