agentleFS
nuncaeslupusnuncaeslupus/literary-critic/.claude/skills/execution/SKILL.md

execution

Skill0 starsChanged 27 days agoMITView on GitHub

Implements code changes from a design, with tests and merge-ready output. Use when the user is building what a design or task describes. Not for investigation (specify), design (design) or one-off scripts.

11 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. 0 are common, 8 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 10 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Keep the notes' Resume section (Decided / Ruled out / Next step) current after RED, after GREEN, and after each decision: it is what the …

    line 17
  2. Distinctive:

    Where the work lives: Run git status --short and check the branch.

    line 31
  3. Distinctive:

    Run every later step from the worktree.

    line 44
  4. Distinctive:

    RED: Before changing production code, write the check that proves the gate and see it fail for the expected reason: a regression test that …

    line 53
  5. Distinctive:

    Read the target code first and match its style, naming and structure.

    line 66
  6. Distinctive:

    When the fix is an API-contract mismatch (arity, kwargs, return shape), pin the contract on the interface itself, so the next caller making …

    line 72
  7. Very specific:

    RECORD: Add the gate evidence to the plan's Evidence log: measured value, exact command, commit SHA and environment provenance (ci, local, …

    line 76
  8. Very specific:

    Run the host repo's lint and the tests for the affected area once (make lint, make test, pytest <path>, …) and fix what fails.

    line 83
  9. Distinctive:

    no debug code, commented-out blocks or stray TODOs.

    line 88
  10. Distinctive:

    no hardcoded secrets, URLs or credentials.

    line 89
  11. Very specific:

    Before opening the PR, run the pre-PR review: one cold reviewer subagent that sees only the packet.

    line 95
  12. Very specific:

    Write the description (linking the ticket or design), add reviewers, and see CI pass.

    line 101
  13. Very specific:

    When arsenal/project/<WORKSPACE>/ exists, read spec.md, plan.md and context.md from there and record gate evidence in that plan.md.

    line 115

“Common” is measured: kept by at least 6 in 10 of the 10 edited versions of this file we could read. “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

Execution Workflow

CANARY: execution-loaded-2026-05-19-4597aff5bb98dd36

Reads status/specification.md and status/plan.md and updates the plan's task statuses as work progresses. Per-task notes go in tmp/<task-id>-notes.md (gitignored); load references/template.md when creating one.

Keep the notes' Resume section (Decided / Ruled out / Next step) current after RED, after GREEN, and after each decision: it is what the session reads back after context compaction.

Work through every task in the plan and on to the PR without pausing between tasks for confirmation; stop only when blocked or before a step that is hard to undo. Deliver the scope the plan asks for, and list pre-existing bugs and unrequested cleanup as follow-ups in the PR description rather than fixing them in this diff.

Steps

Step 1: Prepare

  • The design is approved and the branch is created from the default branch.
  • Where the work lives. Run git status --short and check the branch. A clean tree on the default branch (or on a finished, unrelated branch) works in place. A dirty tree, an in-flight branch the user may still be using, or a long-running process pinned to the checkout gets a worktree, because a worktree never destroys state:

    REPO=$(git rev-parse --show-toplevel)
    TICKET=<ticket-id>
    DEFAULT=$(git -C "$REPO" symbolic-ref --short refs/remotes/origin/HEAD 2>/dev/null || echo origin/main)
    git -C "$REPO" worktree add "../${TICKET}-worktree" -b "${TICKET}-short-description" "$DEFAULT"

    Run every later step from the worktree; remove it with git worktree remove once the PR merges. Ask the user only when the choice is genuinely ambiguous.

  • List the files and tests to create or change, in dependency order, and read the surrounding code for the patterns to follow (naming, structure, error handling).

Step 2: Implement each task — RED → GREEN → RECORD

Each task's Gate in status/plan.md is its measurable acceptance condition.

RED. Before changing production code, write the check that proves the gate and see it fail for the expected reason: a regression test that reproduces the bug, a test that specifies the missing behaviour, or, for a metric gate, the measurement showing the current value misses the threshold. Cover new branches, endpoints and edge cases (null/empty, boundaries, error paths). Name tests test_<what>_<condition>_<expected_result>. Put a new test in a file that is not already the suite's slowest, since the longest file sets the parallel suite's wall clock; load claude-arsenal:core:init § references/performance-tuning.md (§ The long pole) when a test file needs splitting. With <!-- test-discipline: test-after --> in the host CLAUDE.md, write the test alongside the change instead.

GREEN. Implement until the check passes:

  • Read the target code first and match its style, naming and structure.
  • One concern per commit; type hints or strict types; handle errors explicitly; configuration through environment variables or constants.
  • Comments: match the file's density, default to none, and add one only for a why the code cannot show (a hidden invariant, a surprising edge case). Leave task and PR references out of the source.
  • When the fix is an API-contract mismatch (arity, kwargs, return shape), pin the contract on the interface itself, so the next caller making the same mistake is caught too.

RECORD. Add the gate evidence to the plan's Evidence log: measured value, exact command, commit SHA and environment provenance (ci, local, or a project tag). The gate-check skill's run_gate.py --id <task> <measured> reports PASS/FAIL. A task is done when its gate passes and the row is recorded; a plan with no Gate column predates the convention, so skip the record there.

Step 3: Check

Run the host repo's lint and the tests for the affected area once (make lint, make test, pytest <path>, …) and fix what fails; this is the real check behind reporting the code done. When a check cannot run here, say which and why. Refactor with the green suite as the safety net, then confirm:

  • no debug code, commented-out blocks or stray TODOs;
  • no hardcoded secrets, URLs or credentials;
  • the diff matches the design scope;
  • the PR description says what changed, why, and how to test it.

Step 4: Independent review

Before opening the PR, run the pre-PR review: one cold reviewer subagent that sees only the packet. Protocol, exits and round cap: claude-arsenal:core:init § references/pre-pr-review.md.

Step 5: Create the PR

Write the description (linking the ticket or design), add reviewers, and see CI pass. For several PRs that merge in sequence, follow the github skill's stacking reference.


Abbreviation

Abbreviated execution = Steps 2, 3 and 4, where the host repo's CLAUDE.md allows it. Tests, gate evidence and the independent review stay, because abbreviating skips ceremony, not the checks the author cannot do on their own work.

Workspace-aware paths

When arsenal/project/<WORKSPACE>/ exists, read spec.md, plan.md and context.md from there and record gate evidence in that plan.md. Under the task queue, the claimed task's payload at arsenal/tasks/<id>.md carries the acceptance gate, which the queue's gate runner executes before the task is released.

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.