agentleFS
niksavisniksavis/basicly/.agents/skills/harness-client/SKILL.md

harness-client

Skill2 starsChanged 6 days agoMITView on GitHub

Attaches to a running supervisor: shows what the factory does and records answers to its decisions. Use while a supervisor runs or a lane waits on a human judgment.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 4 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    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:

    Every command in this skill is a read or a write to one decision item, except the board's action route.

    line 27
  2. Distinctive:

    If no human is available, leave the item pending or delegate it:

    line 53
  3. Distinctive:

    When the corpus has no answer, it gives the decision back to the human.

    line 60
  4. Very specific:

    do not start a second basicly loop supervise.

    line 68
  5. Very specific:

    do not run basicly loop advance on one of its lanes.

    line 69
  6. Very specific:

    do not delete or edit .basicly/usage/supervisor.lock.

    line 70
  7. Distinctive:

    When the heartbeat is stale, take over with a new supervisor.

    line 73
  8. Distinctive:

    Do not delete the lock file by hand, because another contender can be in the middle of a takeover.

    line 79

≈ “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
<!-- Generated by `basicly skills-build` from skill.yaml. Do not edit; edit the source. -->

Harness Client

This skill makes you a second session that attaches to a running supervisor, observes it and records human answers. To drive work yourself, use harness-loop.

Rules

  • IMPORTANT: never answer a decision for the human. The engine queues only facts that recorded state cannot supply.
  • Record the human's answer word for word. The lane's next dispatch is built from it.
  • Never take the lock from a live supervisor. Two supervisors would land to the same base checkout.
  • Do not approve checkpoints, grant autonomy, land lanes or close issues. Those stay with the session that drives the loop.
  • Poll no faster than the heartbeat. A faster poll gives no newer state and loads the tracker.

Attach

basicly loop session <root-issue>            # live status
basicly loop session <root-issue> --json     # machine-readable

Every command in this skill is a read or a write to one decision item, except the board's action route. loop session reads the tracker and the supervisor lock and writes nothing. Any number of clients can attach. It shows four things:

FieldMeaning
supervisorWho holds the lock and the age of its heartbeat. (none running) means nobody supervises the root. A stale heartbeat means the holder crashed.
laneEach lane in flight, its worktree and branch, if the worktree exists, and its last run (agent, outcome, tokens).
decisionsHow many judgments the session waits on.
grantThe autonomy level in force and the tokens spent against its budget.

The exit status is 0 in all cases. For an exit code that means "blocked on a human", use basicly loop decisions.

Answer a pending decision

basicly loop decisions <root-issue>                            # the queue; exits 1 when an item is pending
basicly loop watch <root-issue>                                # print new pending items
basicly loop watch <root-issue> --once                         # one pass, then exit
basicly loop answer <decision-id> "<answer>"                   # record one answer
basicly loop answer <decision-id> "<answer>" --by "<who>"      # attribute an answerer that is not a human

Show each item to the human as it is: the issue it blocks, its kind, the question and the lane's evidence. The supervisor reads the answer on its next pass.

If no human is available, leave the item pending or delegate it:

basicly loop decide <decision-id> --root <root-issue>

The decider agent answers only from the session's intake corpus, with limited tools. When the corpus has no answer, it gives the decision back to the human. That is the correct result. Do not retry it.

loop watch --interval defaults to the 15-second supervisor heartbeat.

Lock takeover

While loop session shows a fresh heartbeat:

  • do not start a second basicly loop supervise;
  • do not run basicly loop advance on one of its lanes;
  • do not delete or edit .basicly/usage/supervisor.lock.

Answer the pending decision instead. That unblocks the session. When the heartbeat is stale, take over with a new supervisor. It claims the stale lock atomically:

basicly loop supervise <root-issue>

Do not delete the lock file by hand, because another contender can be in the middle of a takeover.

Wall display

To show the board on a shared screen with basicly board serve, read wall display. Use --no-actions for a screen that anyone in the room can touch.

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.