agentleFS
balanced7balanced7/akashic-aurora/.agents/skills/plan-with-the-corpus/SKILL.md

plan-with-the-corpus

Skill8 starsChanged 40 days agoApache-2.0View on GitHub

Use when designing an implementation plan, scoping a slice, starting any multi-step task, or before writing significant code — and use it BEFORE the first Edit, not after. Planning is the highest-leverage review point: a bad plan line becomes hundreds of bad code lines.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 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.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Check for: prior attempts at this exact thing, anti-patterns on the components you'll touch, open blockers/notes that change the premise.

    line 17
  2. Distinctive:

    Read the plan adversarially once: which step hides the risk?

    line 30
  3. Distinctive:

    Keep context lean: hand file PATHS to subagents, never paste bodies ("everything you paste stays resident for the rest of the session").

    line 35
  4. Distinctive:

    When a step's assumption breaks, go back to the plan — don't improvise the architecture mid-edit (see root-cause-before-fixes for the …

    line 39

≈ “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

Plan with the corpus

Review leverage runs research > plan > code. Lessons that arrive while you are editing can only save the line in front of you; lessons pulled while PLANNING save the whole approach. So pull first, plan small, review the plan, then code.

1. Pull before you plan

py agent_cli.py boot <id> --task "<the slice>"     # ranked against the task
py agent_cli.py recall "<topic keywords>"           # targeted corpus search

Check for: prior attempts at this exact thing, anti-patterns on the components you'll touch, open blockers/notes that change the premise. (Plan-time recall also arrives automatically on your prompt when the hook is wired — treat it as a starting set, not the whole corpus.)

2. Plan in verifiable steps

  • Steps sized minutes-not-hours, each with its own check (a test, a command, a render).
  • No placeholders. "TBD", "similar to step 3", "handle errors properly" are plan failures — resolve them now, while it's cheap.
  • State contraindications: what would make this plan WRONG ("don't do this if X").
  • Name the verification for the whole slice up front (which suite, which dogfood command).

3. Review the plan before the code

Read the plan adversarially once: which step hides the risk? which assumption is untested? A minute here outweighs a review cycle later. For irreversible or outward-facing steps, confirm with the human first.

4. Execution hygiene

  • Keep context lean: hand file PATHS to subagents, never paste bodies ("everything you paste stays resident for the rest of the session").
  • Fresh context per phase where natural (research → plan → implement), carrying only the distilled artifact forward.
  • When a step's assumption breaks, go back to the plan — don't improvise the architecture mid-edit (see root-cause-before-fixes for the 3-strikes rule).

5. Close the loop

The plan's surprises are lessons: what the corpus SHOULD have told you before you started is exactly what you record when you finish (learn with a trigger-phrased recommendation), so the next plan starts smarter than this one did.

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.