Execute
bathe your agents in engineering rigour and flames
npx -y skills add davidlee/doctrine --skill executeAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Use to implement one planned, approved phase of a slice — once its design and plan exist and its runtime phase sheet is ready. Move the phase to in_progress, build it TDD red/green/refactor, keep notes current, end green, and surface blockers early.
SKILL.md
6.1 KB, as published. Nobody here has run it
Execute
You are executing one phase of planned work.
Inputs:
- the active runtime phase sheet (
state/.../phases/phase-NN.md) plan.toml/plan.mddesign.md(canonical design reference)slice-nnn.md(scope)
Process
- Confirm the phase's entrance criteria (
EN-) are met for the active phase. - Read
design.md+plan.toml+ the runtime phase sheet before coding. If the sheet is not yet filled, run/phase-planfirst; use/preflightif confirmed inputs, assumptions, and tensions are not yet surfaced. - Identify the concrete files or components you expect to touch first and run
/retrieve-memoryagainst those paths before deep reading or editing, so any scope-bound gotchas or patterns surface early. - Ensure status matches reality before implementation proceeds:
- first phase starting → move the slice:
doctrine slice status <id> started(bare number), if not already there. - flip the phase to
in_progresswithdoctrine slice phase(seeusing-doctrine.md). This auto-records the conformance boundary — the handler stampscode_start_oid = HEADinto the phase sheet. No extra call; the binding rides the transition you already issue (design D5). (It self-skips in a dispatch coordination context, where the funnel beat is the recorder instead.)
- first phase starting → move the slice:
- Implement phase tasks in small coherent units, TDD red/green/refactor write a failing test, make it pass, then refactor. Test behaviour, not trivial implementation. Build and improve test helpers and fixtures as you go.
- After each meaningful unit, run
/harvest. - If that unit produced a durable gotcha, pattern, or subsystem fact worth
future retrieval, run
/record-memorybefore moving on. - Lint as you go (
cargo clippy, zero warnings) and keep the tree buildable. - Follow the repo's commit policy: frequent, small conventional commits scoped
with the slice id (e.g.
feat(SL-009): …). Bias toward a clean tree; don't let.doctrine/**workflow edits drift in a stale uncommitted pile. - If
/preflightor implementation reveals unresolved design ambiguity, unexpected obstacles, tradeoffs, or policy ambiguity, stop and/consultbefore improvising past it. - Keep the runtime phase sheet current as work progresses — never record progress in
authored
plan.toml/plan.md(the storage rule). - Before declaring the phase ready, run the verification gate —
doctrine check gate(lint + test + format) — and review the touched subsystems and notes once more for missed memory-capture candidates. At phase wrap, weigh a per-phase review per the code-review skill's## Cadence— default on below the adherence bar, mandatory on any tripwire. - When exit criteria (
EX-) and verification (VT-, plus any agent/humanVA-/VH-modes) are satisfied, flip the phase tocompletedwithdoctrine slice phase. This closes the conformance boundary — the handler capturescode_end_oid = HEAD, applies the F-6 ancestor/non-merge guard, and upserts the phase's row into the slice's arm-neutral registry. Deterministic, on the critical path — not a "remember to also record" step;slice record-deltais only the manual fallback (correct a range, or bootstrap a pre-binding phase). Then hand off:/phase-planfor the next phase, or — when the slice's phases are all done —doctrine slice status <id> auditand/audit.
Optional: solo isolation (opt-in)
Default execution runs in-tree — the path above is unchanged unless isolation is requested.
Opt-in only — never automatic. Run the phase on its own worktree fork only when the user or the plan explicitly asks for isolation. Absent that annotation, implement in-tree.
When isolation is requested, before implementing (i.e. before step 5) invoke
/worktree with:
mode = solo,allow_work_in_place = true(solo MAY degrade to in-tree on sandbox denial);branch = slice/SL-NNN-slug(the slice id is in scope — e.g.slice/SL-029-dispatch-worktree-creation), worktree dir keyed by the durable id (.worktrees/SL-029).
/worktree handles detection, the creation ladder (doctrine worktree fork),
provisioning, the spawn guards, and the green baseline; the fork branch it
returns is the deliverable.
Assert a clean direct-writer entry. Solo /execute is its own orchestrator —
worker mode is never used here (that is /dispatch's path). Before the TDD
loop, at the solo→direct-writer transition, run:
doctrine worktree status --assert # non-zero `stale-marker` if a stray marker sits here
A stray worker marker in this worktree would make doctrine-mediated writes refuse
mid-work and confuse a direct writer. --assert is exit 0 on a clean entry and
non-zero (stale-marker) otherwise — clear it with doctrine worktree marker --clear --operator before proceeding (bare --clear is refused in a linked
worktree — the §3 accident-fence). (This is the §3 chokepoint the gate PHASE-05 shipped,
now actually called.)
Carry out the TDD loop (steps 5–12) inside the fork. When green, land the fork
onto the coordination branch — /execute is the sole caller of land:
doctrine worktree land --fork slice/SL-NNN-slug # merge --no-ff, ancestry preserved (NEVER squash)
doctrine worktree gc --fork slice/SL-NNN-slug # reap the spent fork once the oracle proves it landed
land preserves the multi-commit TDD history via git merge --no-ff (it cannot
express a squash); gc deletes only after the two-leg landed oracle certifies the
fork (§8.1) — both fail closed with a distinct token, never auto-merge.
Outcomes
- Phase objectives are implemented with traceable evidence, ending green.
- Phase status matches reality during implementation, not only at closure.
- Notes and durable memory stay current throughout execution.