Prd To Spec
Productivityprd-to-specPrd To Spec is an Agent Skill for productivity published in the a5c-ai/babysitter repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Convert an approved PRD into a phase-gated implementation SPEC. User-facing entry point that dispatches to the prd-to-spec babysitter process via /babysitter:call (interactive) or /babysitter:yolo (auto). Stack-agnostic.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/a5c-ai/babysitter/blob/HEAD/library/specializations/product-management/skills/prd-to-spec/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/prd-to-spec, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
PRD to SPEC Generator
User-facing entry point for the PRD → SPEC pipeline. The user runs
/prd-to-spec and this skill dispatches to the companion babysitter
process via /babysitter:call (interactive, with breakpoints) or
/babysitter:yolo (non-interactive, auto-approve).
Usage
/prd-to-spec <PRD-path-or-inline-text>
Examples:
/prd-to-spec docs/active/feature-X-PRD.md/prd-to-spec "Build a notification dashboard with real-time alerts"
Companion process
- Library path:
library/specializations/product-management/prd-to-spec.js - Locally installed:
~/.a5c/processes/prd-to-spec.js
The process file is the source of truth for the pipeline. This skill
intentionally stays thin so users only need to remember /prd-to-spec.
What this skill does when invoked
-
Parse input — resolve the PRD path or capture the inline text.
-
Infer
featureBranch— read the current git branch viagit branch --show-current. If it ismain/develop/dev, ask the user for a feature-branch name. -
Detect optional inputs from the project context:
contextDoc— project conventions doc (defaultCLAUDE.md)archiveDir— past SPEC archive directory (default empty)failureLogPath— project-localfailure-log.md(default empty)secondaryReviewer— hint likecodex,gemini,peer(default empty)
-
Ask the user to pick the run mode:
Run mode for the babysitter
prd-to-specprocess:- Interactive (
/babysitter:call) — recommended. Pauses at every breakpoint (discovery review, SPEC approval, etc.) for explicit user confirmation. - Auto / yolo (
/babysitter:yolo) — non-interactive. Skips all breakpoints; auto-approves every gate. Use only when the pipeline is well-tested and the inputs are trusted.
Default to option 1 if the user does not respond.
- Interactive (
-
Dispatch by invoking the chosen babysitter command via the Skill tool:
For interactive:
Skill('babysitter:call', 'Run the product-management/prd-to-spec process with inputs: prdPath=<resolved>, featureBranch=<resolved>, contextDoc=<resolved>, archiveDir=<resolved>, failureLogPath=<resolved>, secondaryReviewer=<resolved>')For auto / yolo:
Skill('babysitter:yolo', '<same instruction as above>') -
Return the SPEC path and the execution prompt produced by the process to the user.
Fallback (no babysitter runtime)
If neither /babysitter:call nor /babysitter:yolo is available:
- Read the process source at the location above.
- Execute each task's agent prompt manually, in phase order.
- Pause at every
ctx.breakpoint(...)for explicit user approval (interactive equivalent) or skip every breakpoint (yolo equivalent), matching the mode the user picked in step 4. - Return the same outputs the process would have returned.
Optional Integration Hooks
All flow through the process inputs (none required):
- Tracker integration —
trackerHintinput. Examples:jira(/jiraoracli jira workitem update),linear(CLI / MCP),gh-issues(gh issue edit). - Secondary reviewer —
secondaryReviewerinput. Examples:codex(/codex:review),gemini(/gemini-review),deep-verify-plan(/deep-verify-plan),peer(notify a teammate). - Pipeline framework for Phase 3 verification — examples: Dagster, Airflow, dbt, Prefect, K8s preview.
- UI framework for Phase 3 verification — examples: Vite, Next, Storybook.
The process works end-to-end without any of them.
Failure Log Mechanism
The process reads a project-local failure-log.md in Phase 1 to
internalize lessons from prior failed runs. Setup:
- Cross-project lessons:
~/.claude/skills/<your-namespace>/failure-log.md - Project-specific lessons:
<repo-root>/.claude/failure-log.md
Each entry follows:
### [YYYY-MM-DD] <short-pattern-name>
**Context:** <what we tried to do>
**What went wrong:** <the failure mode in 1-2 sentences>
**Constraint going forward:** <the binding rule for future runs>
If the file doesn't exist, the process notes "no prior failures recorded" and continues.
Pre-Push Personal Docs Check
The process executes this gate as part of Phase 7 (Deliver) of the generated SPEC. Procedure:
- List candidates:
git diff --name-only origin/<base-branch>...HEAD - Classify each file:
legitimate/personal-doc/ambiguous - Block by default for:
SPEC*.md,PRD*.md,HANDOFF*.md,SUMMARY*.md,*-NOTES.md, anything underdocs/plans/,ai_docs/,.claude/scratch/,tmp/,investigation-*,analysis-*,playbook-*, AI-generated plans/specs not explicitly approved. - Block-and-ask gate: per-file decision of
unstage/keep/gitignore. Do NOT skip even if rushed. - Proceed with push only after all blocked files are resolved.
Rules
- The skill itself does not produce the SPEC. It only sets up inputs
and dispatches to the process via
/babysitter:callor/babysitter:yolo. - Default to
/babysitter:call(interactive) when in doubt. - Never bypass user-approval breakpoints in interactive mode — they are first-class in the process.
- The SPEC must be proportional to the task (the process honors this).