agentleFS
monaccodemonaccode/astromesh-sdd/AGENTS.md

astromesh-sdd

AGENTS.md0 starsChanged 29 days agoApache-2.0View on GitHub

This project builds features with Spec-Driven Development. Work flows spec → plan → tasks → implement → verify, and every phase leaves a file. Read those files to know where a feature stands; never keep its state only in the conversation. Full skills and templates: https://github.com/monaccode/astromesh-sdd Settings, all optional: .sdd/config.yaml (specs_dir, language, ids, gates, verify_command, phases), .sdd/templates/<name>.md (replace a template), .sdd/rules/all.md and .sdd/rules/<phase>.md (extra rules that win on conflict). Missing files are normal: defaults are specs/, English, REQ/T, gates after…

  • 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 31,774 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 repository7 other files it gives its agents.

The file

We split this AGENTS.md into the rules it gives an agent

4 rules. About 0 are common, 0 are distinctive, and 4 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.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Write approach, decisions (with the rejected alternative), file map, ordered steps each citing REQ-n, risks, test strategy per requirement, …

    line 46
  2. Very specific:

    Verify: Run verify_command now and keep the output tail.

    line 59
  3. Very specific:

    Gate: after each phase in gates, show the artifact and stop until the user approves.

    line 65
  4. Very specific:

    Traceability: every step and task cites a REQ-n.

    line 67

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

Spec-Driven Development (astromesh-sdd)

This project builds features with Spec-Driven Development. Work flows spec → plan → tasks → implement → verify, and every phase leaves a file. Read those files to know where a feature stands; never keep its state only in the conversation. Full skills and templates: https://github.com/monaccode/astromesh-sdd

Artifacts

specs/
  constitution.md          project principles (optional)
  NNN-<slug>/
    spec.md                REQ-n requirements: what and why, testable
    plan.md                decisions, file map, steps citing REQ-n
    tasks.md               T-n checklist, each with a runnable check
    verify.md              REQ → task → evidence; verdict PASS/FAIL

Settings, all optional: .sdd/config.yaml (specs_dir, language, ids, gates, verify_command, phases), .sdd/templates/<name>.md (replace a template), .sdd/rules/all.md and .sdd/rules/<phase>.md (extra rules that win on conflict). Missing files are normal: defaults are specs/, English, REQ/T, gates after specify, plan and tasks.

Where am I?

In the feature dir (the one named, else the highest NNN), walk phases in order and do the first phase that is not done: no spec.md → specify; no plan.md → plan; no tasks.md → tasks; unticked tasks → implement; no verify.md → verify; verify.md PASS → done. A gated artifact whose **Status:** is not approved sends you back to its gate. verify.md FAIL → do its Next actions, then run sdd-verify again.

Phases

Constitution (optional, once). Read README, build files, CI; write 5–9 testable principles with a one-line why, plus the real test and lint commands.

Specify. Next free NNN, kebab slug. Ask at most 3 scope questions, one at a time; anything else unknown becomes [NEEDS CLARIFICATION: …] — never guess. Each REQ-n is one observable behavior with Given/When/Then acceptance. List non-goals. No implementation details. To change a spec: new requirements take the next free ID; IDs never renumber and are never reused; removed ones become ~~REQ-n~~ (removed: reason); log it in Changelog.

Plan. Resolve open questions first. Trace the real code path the change touches. Write approach, decisions (with the rejected alternative), file map, ordered steps each citing REQ-n, risks, test strategy per requirement, and a coverage table — a requirement no step cites blocks the plan.

Tasks. One reviewable unit per task, tests in the same task as their code: - [ ] T-n — title (REQ-a) with files:, check: (runnable command → expected), depends:. Mark parallel tasks [P]. Coverage table: every requirement appears.

Implement. One task at a time: write the test, see it fail for the right reason, implement the minimum, run the task's check and the suite, tick only on pass (— done YYYY-MM-DD), commit citing T-n. A new or changed requirement goes to the spec first.

Verify. Run verify_command now and keep the output tail. For each requirement: the tasks citing it, evidence produced now, ✅ / ⚠️ / ❌. PASS only if the suite passed, every requirement is ✅ and no task is unticked. Write verify.md; on FAIL, list where to go back.

Rules in every phase

  • Gate: after each phase in gates, show the artifact and stop until the user approves; then set its **Status:** line to approved. Approval lives in the file.
  • Traceability: every step and task cites a REQ-n; uncited requirements and unciting tasks are failures.
  • Spec first: requirements change in spec.md before code changes.
  • Evidence: "done" and ✅ need a check that ran in this session.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.