agentleFS
albiol2004albiol2004/trio-agent-loop/.claude/skills/trio-init/SKILL.md

trio-init

Skill2 starsChanged 26 days agoMITView on GitHub

Initialize the trio agent loop in the current project — creates the loop/ mailbox directory and GOAL.md from the given goal text.

The file

We split this skill 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
  1. Very specific:

    Mailbox directory: default loop/. If $ARGUMENTS starts with dir=<path> (e.g.

    line 10
  2. Distinctive:

    Create the mailbox directory with these files:

    line 13
  3. Very specific:

    If the project is a git repo, ask the user whether loop/ should be gitignored (ephemeral) or committed (auditable history).

    line 44

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill 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
SKILL.md

Set up the trio-loop mailbox in the current project.

Mailbox directory: default loop/. If $ARGUMENTS starts with dir=<path> (e.g. /trio-init dir=loop-authz add rate limiting…), use that directory instead — every loop/ reference below then means that directory. One mailbox per loop is a hard rule: two loops (or two sessions) sharing a mailbox corrupt each other's state.

  1. If the mailbox's STATE.md already exists, show its current status and LOG.md tail, then ask the user to choose: (a) reset — archive the old mailbox to loop-archive-<date>/ and start fresh; (b) abort; or (c) keep it and initialize this new loop in a sibling directory (loop-<slug>/ from the new goal) — the right choice when the existing loop may still be running in another session. If STATE.md says status: running and LOG.md's last entry is recent, recommend (c) — never hijack a possibly-live mailbox.
  2. Create the mailbox directory with these files:

loop/GOAL.md — from $ARGUMENTS. If arguments are empty or vaguer than one sentence of substance, interview the user briefly (what does done look like, hard constraints, what must NOT change) before writing it. Decide the profile: software (default) or data — choose data when the goal is pipelines, ETL/ELT, SQL models, notebooks, finance/reporting datasets. Structure:

# Goal
profile: software | data
<the mission, in the user's words, sharpened>
## Definition of done
<objectively checkable statements>
## Constraints
<stack, style, things that must not break, budget notes>

For profile: data, the Definition of done MUST include data ground truth, so interview for it if missing: source(s) of truth to reconcile against, tolerated row-count/aggregate deltas, key uniqueness expectations, and whether re-runs must be idempotent. Vague data goals ("clean the dataset") are the #1 cause of runaway loops — pin numbers down.

loop/STATE.md (the mission: line is the first sentence of GOAL.md's mission, verbatim — the /trio orchestrator halts if it ever stops matching GOAL.md, which catches another session repurposing the mailbox):

schema: 1
iteration: 0
max_iterations: 10
status: ready
mission: <first sentence of the goal, verbatim>

## Approaches tried and rejected
(append-only; the Lead adds one line per dead end, with why)

## Key decisions and rationale
(append-only)

loop/LOG.md with a # Trio loop log header line, and empty loop/PLAN.md, loop/REPORT.md, loop/VERDICT.md.

  1. If the project is a git repo, ask the user whether loop/ should be gitignored (ephemeral) or committed (auditable history). Default suggestion: commit it.
  2. Confirm setup and print next steps: run a Claude trio loop — ask Claude to "run the trio loop" (the trio-claude skill drives the release launcher ~/.local/share/trio-agent-loop/current/native/launch.sh start: host engine, frozen acceptance per profile, root-free, lands at SHIP). /trio runs a single supervised in-session iteration for deliberate step-through (/loop /trio is only the legacy in-session step-through, not the recommended autonomous path). If the mailbox is not loop/, say to name it: "run the trio loop on <path>" or /trio dir=<path>. Mention that GOAL.md and max_iterations in STATE.md are the two human control knobs while a loop runs.

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.