From skillry-product-docs-and-research
Use when you need to write strict final execution prompts for Codex, Claude, Copilot, or other coding agents.
How this skill is triggered — by the user, by Claude, or both
Slash command
/skillry-product-docs-and-research:64-codex-final-execution-promptThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Write strict final execution prompts for Codex, Claude, Copilot, or other coding agents. A final execution prompt has one singular, bounded goal; every instruction in it is testable; and a stop condition exists for every ambiguous or blocked state so the agent never silently guesses. The output is a self-contained prompt — exact context, exact changes, hard constraints, verification commands, a...
Write strict final execution prompts for Codex, Claude, Copilot, or other coding agents. A final execution prompt has one singular, bounded goal; every instruction in it is testable; and a stop condition exists for every ambiguous or blocked state so the agent never silently guesses. The output is a self-contained prompt — exact context, exact changes, hard constraints, verification commands, acceptance checklist, output contract, and stop conditions — ready to paste into a coding agent.
04-implementation-plan) and now needs translating into a strict agent-executable format with exact file/change pairs and verification commands.63-md-spec-generator for human-facing specs and handoff packs, 04-implementation-plan for the planning phase that precedes prompt-writing.tsc --noEmit → 0 errors).git push --force, no DROP, no new dependencies).# Task: <single, specific goal>
## Context
- Repo/paths: <...> PM: <npm/pnpm/yarn/...> Framework/runtime: <...>
- Relevant files: <path:line ...>
## Do exactly this
1. <change> in <file>
2. <change> in <file>
## Constraints (hard)
- Do NOT: deploy, run prod migrations, reset/seed data, force push,
add dependencies, or touch <off-limits paths>.
- Preserve: <public API / behavior / existing patterns>.
- Match existing style and conventions.
## Verify (must pass before done)
- `<typecheck cmd>` → clean (0 errors)
- `<test cmd>` → green
- `<build cmd>` → succeeds
## Acceptance criteria
- [ ] <observable outcome>
- [ ] <observable outcome>
## Report back
- Files changed (path → what), commands run + results,
residual risks, anything skipped + why.
## Stop conditions
- If a constraint blocks the task, a needed secret/decision is missing,
or a verification command fails after one fix attempt: STOP and report
instead of guessing or widening scope.
A risk-tier example for a rename task, showing the forbidden list doing real work:
## Constraints (hard) — rename `getUser` → `fetchUser`
- Rename only within `src/`; do NOT touch `node_modules`, generated files, or `*.snap`.
- Update all call sites; do NOT change the function body or signature.
- Do NOT add a backward-compat alias unless a call site is outside this repo.
A vague ask ("add validation to the signup endpoint") becomes a final execution prompt:
# Task: Reject signup requests with an invalid email at api/signup.ts
## Context
- Repo/paths: api/signup.ts, api/__tests__/signup.test.ts
- PM: pnpm Framework/runtime: Express + TypeScript + Zod (already a dependency)
- Relevant files: api/signup.ts:18 (handler), api/schema.ts (Zod schemas live here)
## Do exactly this
1. In api/schema.ts, add `signupSchema = z.object({ email: z.string().email(), password: z.string().min(8) })`.
2. In api/signup.ts:18, parse the body with `signupSchema.safeParse(req.body)`;
on failure return 400 with `{ error: 'invalid_input' }` before any DB call.
## Constraints (hard)
- Do NOT: deploy, run migrations, add a new dependency (Zod is already present),
change the success response shape, or touch any file outside the two listed.
- Preserve: the 201 success response and its body.
- Match existing style (named exports, no default export).
## Verify (must pass before done)
- `pnpm tsc --noEmit` → 0 errors
- `pnpm test api/signup` → green (add cases: bad email, short password)
- `pnpm lint api/signup.ts` → clean
## Acceptance criteria
- [ ] POST with a malformed email returns 400 and never reaches the DB.
- [ ] POST with a valid body still returns 201 with the unchanged shape.
## Report back
- Files changed (path → what), commands run + results, residual risks.
## Stop conditions
- If Zod is NOT already a dependency, or a verify command fails after one fix
attempt: STOP and report. Do not install packages or widen scope.
Note how the forbidden list ("Zod is already present", "do not touch other files") and the stop condition do the real work: they prevent the agent from npm install-ing a second validation library or "improving" adjacent code.
npm install its way out of a type error.Return the finished execution prompt using the template, ready to paste into a coding agent. It must be self-contained: exact context, exact changes, hard constraints, verification commands, acceptance checklist, output contract, and stop conditions — with zero ambiguity.
Done means the prompt pins context, lists exact changes plus hard constraints, specifies exact verification and acceptance criteria, defines the report-back contract and stop conditions, and contains no ambiguous instruction.
Provides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.
npx claudepluginhub fluxonlab/skillry --plugin skillry-product-docs-and-research