agentleFS
stormlightlabsstormlightlabs/mire/.claude/skills/arclightning/SKILL.md

arclightning

Skill1 starsChanged 19 days agoApache-2.0View on GitHub

Pick up, work, hand off, and complete tasks tracked by Arc Lightning (`arcl`). Use in a project with an `.arcl/` directory, when asked what to work on next, to start or finish an arcl task, or to leave a resume note for the next session.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 2 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

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

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

    When the arcl-mcp server is connected, call its tools instead of the commands.

    line 13
  2. Very specific:

    Every --json command writes one object to standard output:

    line 20
  3. Very specific:

    Start it. arcl --json task start <id> [task_start] moves the task to in progress.

    line 43
  4. Very specific:

    Do not pass --force unless the person asks.

    line 46
  5. Very specific:

    Read the context. arcl --json context <id> [context] returns the task with its ancestors, spec, plan, phase, blockers, dependents, …

    line 47
  6. Distinctive:

    Do the work the task describes.

    line 51
  7. Very specific:

    Use --evidence-file - and --note-file - to pass longer Markdown on standard input.

    line 63
  8. Distinctive:

    When you find work outside the task, capture it instead of doing it:

    line 68
  9. Very specific:

    Run arcl <command> --help for every flag, and arcl check to validate the project.

    line 80

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Arc Lightning

Arc Lightning tracks a project's tasks in .arcl/. Drive it through the arcl CLI with --json, and read fields from the envelope rather than from terminal text. arcl decides readiness and which transitions are allowed. Ask it instead of working that out yourself.

When the arcl-mcp server is connected, call its tools instead of the commands. Each step below names the tool in brackets. A tool returns the same envelope as its command, as structured content. Its input schema lists the fields it takes, which do not always match the command's options.

Read the envelope

Every --json command writes one object to standard output:

{ "format_version": <n>, "data": { "...": "..." } }

The version depends on the command. Check it before reading data. A failure goes to standard error as {"format_version": 1, "error": {"code", "exit_code", "message"}}, and the command exits with that nonzero exit_code. A failed tool call returns the same object with isError set. The automation guide lists the version each command returns, the readiness reason kinds, and their ID fields.

Work the loop

  1. Find work. First resume your own parked work: arcl --json list --kind task --status parked [list] lists tasks a session handed off. Otherwise arcl --json ready [ready] returns data.tasks in order, and arcl --json next returns the first as data.task. An empty list means nothing can start. The tools return at most limit records and a data.next_cursor to pass as cursor while more remain.
  2. Start it. arcl --json task start <id> [task_start] moves the task to in progress. If it refuses with invalid_request, run arcl --json explain <id> [explain] and read data.readiness.reasons. Do not pass --force unless the person asks. The tool has no force.
  3. Read the context. arcl --json context <id> [context] returns the task with its ancestors, spec, plan, phase, blockers, dependents, readiness, and earlier evidence. data.task.handoff holds the last resume note, and data.task.body holds the acceptance criteria.
  4. Do the work the task describes.
  5. Finish or hand off:
    • Done: arcl --json task complete <id> --evidence "<proof>" [task_complete], where the evidence says what shows the work is done. data.unblocked lists the tasks this made ready. Start the next one from there or go back to step 1.
    • Not done: arcl --json task handoff <id> --note "<note>" [task_handoff], where the note says where the work stands and what is next. This parks the task. The next session finds it in step 1, runs arcl --json task unpark <id> [task_unpark], then starts it again and reads the note from the context.

Use --evidence-file - and --note-file - to pass longer Markdown on standard input.

Record new work

When you find work outside the task, capture it instead of doing it:

arcl --json capture create "<title>" --body "<what and why>"

The tool is capture_create. The server also has capture_promote, show, spec_create, plan_create, task_create, task_update and note_create, for arcl capture promote, arcl show, arcl spec create, arcl plan create, arcl task create, arcl task update and arcl note create.

Run arcl <command> --help for every flag, and arcl check to validate the project.

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.