agentleFS
hmj1026hmj1026/dhpk/skills/flow-drive/SKILL.md

flow-drive

Skill2 starsChanged 7 days agoMITView on GitHub

Use the current Host and Model to complete one bounded, authorized task and report observed verification evidence or a precise blocker.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is 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.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Very specific:

    Follow the consumer repository's own instructions and Project Convention.

    line 27
  2. Distinctive:

    Verify the acceptance criteria using the consumer project's appropriate tools.

    line 29
  3. Distinctive:

    Do not claim a commit, merge, release, deployment, or other external action unless it was separately requested and actually completed.

    line 32

≈ “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

Flow Drive

Invoke $flow-drive explicitly with one bounded task, a task-file path, or a confirmed specification/change ID. The current Host and Model handle understanding, investigation, edits, verification, and the report. $flow-drive --help describes the current entry; help makes no project changes.

This directory is the complete skill. It needs no dhpk checkout, sibling skill, resource synchronization, parser, runner, external CLI worker, or special execution receipt. The Host must be able to read the supplied task and relevant project files, make only authorized edits, and run the consumer project's available verification. It uses its normal file and shell tools; provider and subagent access remains subject to that Host's own capabilities and grants.

Invocation boundary

Accept one task input and no execution-selection options. If a user supplies a retired Flow Drive role, backend, Provider, planning, or architecture option as invocation syntax, reject it before any dependent edits, explain that it is unsupported here, and point to the migration guide. Ask for one task-only input or a separate help invocation. Do not silently reinterpret a retired option as task text.

A flag written inside the quoted task, a task file, or the confirmed specification is task content. For example, $flow-drive "Update the documentation example that contains --worker=codex" is an ordinary documentation task. Read and interpret that content as part of the task; do not treat the embedded flag as a Flow Drive option.

Complete the task

  1. Confirm one bounded goal, its acceptance criteria, requested scope, and the user's authority. If a confirmed ID or task file cannot be read, or the outcome or write scope is missing, state the exact blocker before dependent edits.
  2. Follow the consumer repository's own instructions and Project Convention. Do not impose dhpk development or release rules on a consumer project.
  3. Investigate enough to support the requested change. Diagnose an unknown cause before edits that depend on it. Keep writes within the authorized scope and preserve unrelated work.
  4. Verify the acceptance criteria using the consumer project's appropriate tools. Report the commands or checks actually run and their observed results. A failed or unavailable check is a failure or gap, never a pass.
  5. Finish when every criterion has evidence, or report what remains blocked or unverified and the concrete next action.

Do not claim a commit, merge, release, deployment, or other external action unless it was separately requested and actually completed.

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.