pacer
AGENTS.md1 starsChanged 21 days agoApache-2.0View on GitHub
This repository is public. Agent working files stay out of git. These locations override the docs/superpowers/... defaults in the superpowers skills. Do not commit anything under .work/. Summarize the plan in the PR description instead. Tracked files must not contain machine-specific paths (/Users/..., /Volumes/...) or personal emails. Test fixtures that are needed go in test/fixtures/ and use $HOME or temp-dir paths. The machine you work on runs the maintainer's own pacer. Development, tests and live validation must not change it,…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 33,315 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository6 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- project-governanceSkill
- project-reviewSkill
- live-validateSkill
- project-governanceSkill
- project-reviewSkill
The file
We split this AGENTS.md into the rules it gives an agent
3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 16
Do not commit anything under .work/.
- Very specific: line 19
Test fixtures that are needed go in test/fixtures/ and use $HOME or temp-dir paths.
- Very specific: line 39
Before working, read .agents/skills/project-governance/SKILL.md, .project/constitution.md and .project/context.md.
≈ “Common” is an estimate: at least 1 in 5 of the 42,283 AGENTS.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
Agent conventions
Where working documents go
This repository is public. Agent working files stay out of git.
| Document | Location | Tracked |
Implementation plans (superpowers writing-plans) | .work/plans/YYYY-MM-DD-<feature>.md | no |
| Code reviews, verdicts, task briefs | .work/reviews/ | no |
| Transcripts, captured tool output, scratch fixtures | .work/fixtures/ | no |
Design specs (superpowers brainstorming) | docs/specs/YYYY-MM-DD-<topic>-design.md | yes |
| Architecture decisions (ADRs) | docs/decisions/NNNN-<title>.md | yes |
These locations override the docs/superpowers/... defaults in the superpowers skills.
Do not commit anything under .work/. Summarize the plan in the PR description instead.
Tracked files must not contain machine-specific paths (/Users/..., /Volumes/...) or personal emails.
Test fixtures that are needed go in test/fixtures/ and use $HOME or temp-dir paths.
Never touch the user's real pacer install
The machine you work on runs the maintainer's own pacer. Development, tests and live validation must not change it, including with ad hoc shell commands. Never write, copy, link, sign in place, replace, launch or delete:
/Applications/pacer.app(or anything inside it), including backup or nested copies next to it;~/.local/bin/pacerand~/.local/bin/pacer-core;- the default data home
~/.pacer; - the
com.pacer.serveservice: its plist in~/Library/LaunchAgentsor the launchd label. Neverlaunchctl bootstrap,bootoutorkickstartit, and never load your own jobs that run the real binaries against the real home.
Use a temporary PACER_HOME and HOME, the fixture bundle dev.pacer.desktop.fixture in a temporary app directory, a separate port, and binaries under your worktree (dist/, target/). Contents/MacOS/pacer is the desktop executable, never the CLI; the CLI pair is Contents/Resources/pacer and pacer-core (see ADR 0021). Scripts enforce this with scripts/lib/real-install-guard.ts. Set PACER_ALLOW_REAL_INSTALL=1 only when the maintainer asks you to change their install. Unload any launchd job and stop any process you started before you finish.
Commits
Conventional Commits (feat:, fix:, docs:, chore:, test:). PRs are squash-merged, so the PR title becomes the commit on main.
Sign off every commit (git commit -s); the DCO check rejects PRs with unsigned commits. See CONTRIBUTING.md.
Project context
Before working, read .agents/skills/project-governance/SKILL.md, .project/constitution.md and .project/context.md. Repository-local skills are tracked and require no global installation. If your agent does not discover .agents/skills/, read the referenced files directly; Claude adapters under .claude/skills/ point to the same instructions.
From the context index, load the scoped rules (.project/rules/) and reviewer definitions (.project/reviewers/) that match the areas your task touches. Pass these references and the applicable rule IDs to any workflow you select and to any reviewer you delegate to. Keep the project's established constraints in effect through planning, implementation and review. Surface conflicts and unresolved decisions (.project/decisions.md). Change a rule only through the amendment process in the constitution.
Agent work routing
If you are the planner, follow REPO-11 to REPO-14 in .project/rules/repository.md (proposed, amendment 0.15.0). In short, delegate execution through pacer as the current mode requires, pick the smallest model tier that can do the job for any work on your side, and record the choice in each brief and report. Executors don't delegate further except through an API-5 grant.
Definition of done
A task that changes runtime behavior is done only after it passes live validation with the harness (bun run live). Follow docs/testing/live-validation.md and include the evidence report's summary in the result (rule CT-11). Report surfaces you could not exercise as UNVERIFIED.
Before delivering changed code as complete, follow .agents/skills/project-review/SKILL.md: obtain independent project review, fix confirmed blocking findings, rerun affected validation, and independently review the fixes against the final revision (REPO-10). Store reports under .work/reviews/ and summarize the evidence in the result and PR. Documentation and skill changes also require a proportionate repository-rule review. Missing independent review or unresolved required checks mean UNVERIFIED/incomplete, not 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.

