promptdriven/pdd

pdd-small-change

PDD small-change workflow.

Quelltext ansehen

Verwendung

  1. Installationsbefehl kopieren und das Quellpaket vor der Installation in der eigenen Agentenumgebung prüfen.
  2. Benötigte Werkzeuge, Zugangsdaten und Hosts im Original prüfen.
  3. Eingabe und gewünschtes Ergebnis beschreiben und die Ausgabe mit den Quellen abgleichen.
npx skills add https://github.com/promptdriven/pdd --skill pdd-small-change

Vor dem Start

  • Die Installation umfasst weder Modellabo noch API-Kontingent oder Dateizugriffsrechte.
  • Sterne und Installationen zeigen Aktivität, keine unabhängige Qualitäts- oder Sicherheitsprüfung.
Originales Skill-Dokument

Aus dem Quell-Repository gerendert; Überschriften, Beispiele, Code, Tabellen, Links und Bilder bleiben erhalten.

PDD small change

Fix both: the brakes (the real defect) and the warning light (a test or check that would catch it next time). Brakes first. If the agent cannot find the defect yet, it fixes the warning light first: it adds what makes the defect visible (a failing reproducer test, a precise error, logging), so the cause becomes easy to find, and then it fixes the brakes. A warning-light-only change must say 'Not fixed yet: this helps find the cause' on the PR, and the issue stays open; it never closes the bug.

Use this strictly for changes outside PDD: docs, guides, package.json, lockfiles, config, CI, metadata and test maintenance — never *.prompt.md, user_stories/*.md, owned files or product code (src/, app/, etc.). Product or prompt-owned code always uses pdd-bug or pdd-feature, even for a small fix or with no prompt yet (create the prompt in PR 1, never bypass PDD).

Each area is one main chat; read its step files.

  1. plan: steps/scope.md → gate: approve
  2. build: steps/change.md, steps/verify.md, steps/formal.md → gate: pr

Rules

  • Main chat coordinates; no code or checks.
  • Fresh subagents in step order once files exist; never reuse builders or let them check code. Step file/paths at once, no chat history. Briefs: "the run's main chat", actual actors/root paths. Wait one long time-out, no small follow-ups; §11, RUNBOOK "Bounded handoffs".
  • Checks: fresh first. Only verify/formal may reuse after startup failure; wrote no code, tests or prompts this run.
  • Read pdd-prompting RUNBOOK §17 before checks: cleanup, unrun checks, recovery and result replacement.
  • At gates, show results, ask and stop. Revise changes and ask again. Outside runs, ask before opening PRs. In runs, pdd opens them. Only the main chat may merge on explicit person request under RUNBOOK §2.
  • Use full URLs.
  • For structural cleanup, follow docs/refactor-runbook.md.
  • Before an optional step, ask the person and wait for the answer. In a run, run <pdd> workflow ask first.
  • A person decides expected values. Never change a test to make it pass.

In a run

  • Write <run folder>/STEP.md; commit changes.
  • For an answer mid-step, run <pdd> workflow ask "<question>". Ask in plain chat, never requestuserinput, and end your turn. The person replies in chat or with <pdd> workflow changes N "answer".
  • Show the gate; judge the later chat reply without exact phrases. Never approve questions or changes. Record gate approval:
sh
<pdd> workflow approve N --said - <<'PDD_SAID'
<the person's exact words>
PDD_SAID

Quote a delimiter absent as a line. Stdin keeps newlines; blank fails. RUNBOOK §2.

aus demselben Repository

Weitere Skills

Alle Skills