skilld
All skills
av avatar

/timeboxed-iterating

@5588d55
by Ivan Charapanauav/skills16 stars
2

Run a task iteratively over a user-specified duration by dispatching subagents. The orchestrator stays strictly linear and time-checked, but each iteration fans out MULTIPLE subagents in parallel over independent units (and those subagents may fan out further). Shared state — a compact progress digest and a growing environment cheatsheet — lives on disk and is passed to every subagent BY REFERENCE, killing the per-subagent rediscovery tax. Every subagent first reads an initialiser preamble that points it at that shared state and makes it write findings back, so the cheatsheet populates itself; role prompts are produced ONCE on the filesystem by a single scaffold command and dispatched by path, never re-typed per dispatch. Subagents persist their results to disk and return only a tiny status, so the orchestrator's context stays small and lasts. The orchestrator classifies the goal as FINITE or OPEN-ENDED: finite lists may finish early; open-ended goals use the full duration, never idle, never manufacture busywork. Use when the user gives a task and a duration and wants it ground out iteratively by subagents over that time.

Ask your Agent

Use this Skill: https://skilld.dev/gh/av/skills/timeboxed-iterating

This session only. Nothing lands on disk.

scaffoldbuilder.md

≈555 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Builder — one unit of a batch

Read {{HARNESS}}/prompts/initialiser.md FIRST — it makes you read the shared state (cheatsheet + digest) and append any reusable finding back before you return. Then do your unit.

You are one subagent in a parallel batch. You own ONE unit of work; other subagents work other units concurrently — stay strictly inside your unit.

Your unit

Your unit id, its scope, and this unit's gap/target are in your dispatch delta and your unit file {{HARNESS}}/units/<unit-id>.md. Edit ONLY the files that file scopes to you; another subagent owns the rest. The overall goal ({{GOAL}}) is in the digest header.

Your task

Do this one unit of real work toward the goal, end to end.

  • If the unit contains independent sub-units, you MAY fan out your own sub-subagents in parallel over them — hand each the PATH {{HARNESS}}/prompts/sub-subagent.md plus its sub-unit delta (do not re-type the template). Size each sub-subagent's model to its sub-unit; never request a model more powerful than the one you are running on.
  • Produce a tangible artifact — committed code, written content, a real finding/fix — NOT a plan, not a list of suggestions, not analysis.
  • Commit before returning: timeboxed(<unit-id>): <summary>. The unit-id prefix is how an interrupted run attributes your commit — never omit it.

Persist to disk, return almost nothing (context longevity)

  • Write your FULL result into {{HARNESS}}/units/<unit-id>.md (your own file — do NOT write to progress.md's body or other units' files; that causes write contention). Files touched, changes made, what remains — all goes HERE.
  • Append any reusable finding to {{HARNESS}}/cheatsheet.md (the initialiser told you to).
  • Append ONE status line to the digest {{HARNESS}}/progress.md under the iteration log, e.g. - <unit-id>: built, commit <hash>, awaiting measurement.

Return (TINY — one structured line, no prose dump)

unit <unit-id>: <done|blocked>, commit <hash> — plus at most ONE short clause if blocked. Do NOT paste your diff, your artifact, or your full write-up into the return; it lives in your unit file. The orchestrator reads the digest tail, not your return body.

Source: SKILL.md on GitHub

Third-party checks

No alerts12d3 checks · Risk SAFE
  • Gen Agent Trust Hub12d

    The skill establishes an autonomous iteration loop that delegates tasks to sub-agents for a fixed duration. While this is the intended design, it creates a surface for indirect prompt injection and resource exhaustion if the sub-agents process untrusted data.

  • Socket12d

    No alerts

  • Snyk12d

    Risk: LOW · No issues

Provenance

Signed by skilld at 5588d55. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub last week.

Activeupdated last month

README badge

README badge for av/skills/timeboxed-iterating