Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Convert a raw task (tracker ticket, email, or text) into a fully characterized PRD. User-facing entry point that dispatches to the task-to-prd babysitter process via /babysitter:call (interactive) or /babysitter:yolo (auto). Stack-agnostic.
.claude/skills/a5c-ai-task-to-prd/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 160% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 84% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 81% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 25% | 0% |
User-facing entry point for the task → PRD pipeline. The user runs /task-to-prd and this skill dispatches to the companion babysitter process via /babysitter:call (interactive — Five Whys, clarification loop, scope-lock, per-finding gates, final approval) or /babysitter:yolo (non-interactive, auto-approve every gate).
/task-to-prd <tracker-id | file-path | inline-text>Examples:
/task-to-prd JIRA-1234/task-to-prd path/to/email.txt/task-to-prd "Eden asked to change the donut to show lessons" (inline)library/specializations/product-management/task-to-prd.js~/.a5c/processes/task-to-prd.jsThe process file is the source of truth for the pipeline. This skill intentionally stays thin so users only need to remember /task-to-prd.
path, or inline text.
featureBranch — read the current git branch viagit branch --show-current. If unsuitable (e.g., main), ask the user to provide a feature-branch name.
contextDoc — project conventions doc (default CLAUDE.md)archiveDir — past PRD archive directory (default empty)trackerHint — hint like jira, linear, gh-issues (default empty)secondaryReviewer — hint like codex, gemini, peer (default empty)> Run mode for the babysitter task-to-prd process: > > 1. Interactive (/babysitter:call) — recommended. The Five > Whys + clarification loop is intrinsically interactive — every > question goes through the user. Verification findings each get > a per-finding approval gate. > 2. Auto / yolo (/babysitter:yolo) — non-interactive. Skips > the clarification loop (uses defaults from Five Whys), auto- > approves every gate, applies all proposed PRD changes > automatically. Use only when the input is already well-defined > and you want a fast first draft.
Default to option 1 if the user does not respond.
For interactive: Skill('babysitter:call', 'Run the product-management/task-to-prd process with inputs: input=<resolved>, featureBranch=<resolved>, contextDoc=<resolved>, archiveDir=<resolved>, trackerHint=<resolved>, secondaryReviewer=<resolved>')
For auto / yolo: Skill('babysitter:yolo', '<same instruction as above>')
/prd-to-spec prompt produced by the process to the user.
If neither /babysitter:call nor /babysitter:yolo is available:
ctx.breakpoint(...) for explicit user approval(interactive equivalent) or skip every breakpoint (yolo equivalent), matching the mode the user picked in step 4.
All flow through the process inputs (none required):
trackerHint input. Examples: Jira(/jira skill or acli jira workitem view), Linear (CLI / MCP), GitHub Issues (gh issue view).
secondaryReviewer input. Examples: Codex(/codex:review), Gemini (/gemini-review), deep-verify-plan (/deep-verify-plan), peer (notify a teammate).
attribution to user or a named stakeholder (e.g., product owner, tech lead, designer).
The process works end-to-end without any of them.
| Phase | What happens | Interactive in /babysitter:call? | |-------|--------------|-----------------------------------| | 1a. Load source | Detect tracker / file / inline | no | | 1b. Five Whys | Root-cause analysis | no | | 1c. Clarification loop | Q&A one at a time | YES | | 1d. Scope Lock | Approve scope before drafting | YES | | 2. Draft (parallel) | Codebase scan + draft PRD via embedded skill instructions | no | | 3a. Verification (5 parallel) | what-could-go-wrong + consistency + conventions + adversarial + quality checklist | no | | 3b. Per-finding gate | Approve each proposed PRD change | YES | | 4a. Final review | Approve final PRD | YES | | 4b. Tracker update | Push reference to tracker (if applicable) | no | | 4c. Follow-up prompt | Output /prd-to-spec <path> | no |
In /babysitter:yolo mode, every gate marked YES is auto-approved.
and dispatches to the process via /babysitter:call or /babysitter:yolo.
/babysitter:call (interactive) when in doubt.effectively turns it into a shallow first-draft using only Five-Whys defaults. Recommend interactive mode for any non-trivial task.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 3,155 | 8,157 | +159% | 1 | 1 | 0% | 437 | 1,916 | +338% | 0 | 0 | — |
case-02 | fail→fail | 12,606 | 5,817 | -54% | 1 | 1 | 0% | 1,667 | 1,867 | +12% | 0 | 0 | — |
case-03 | fail→fail | 15,963 | 6,741 | -58% | 1 | 1 | 0% | 2,783 | 1,978 | -29% | 0 | 0 | — |
case-04 | fail→fail | 32,061 | 6,447 | -80% | 1 | 1 | 0% | 6,201 | 1,827 | -71% | 0 | 0 | — |
case-05 | fail→fail | 9,980 | 5,321 | -47% | 1 | 1 | 0% | 1,692 | 2,237 | +32% | 0 | 0 | — |
case-06 | fail→fail | 34,229 | 35,940 | +5% | 1 | 1 | 0% | 8,217 | 9,705 | +18% | 0 | 0 | — |
case-07 | fail→fail | 5,293 | 3,928 | -26% | 1 | 1 | 0% | 832 | 2,250 | +170% | 0 | 0 | — |
case-08 | fail→pass | 6,618 | 6,601 | -0% | 1 | 1 | 0% | 974 | 2,531 | +160% | 0 | 0 | — |
case-09 | pass→pass | 12,870 | 10,570 | -18% | 1 | 1 | 0% | 2,281 | 3,489 | +53% | 0 | 0 | — |
case-10 | fail→pass | 11,167 | 4,090 | -63% | 1 | 1 | 0% | 2,201 | 2,326 | +6% | 0 | 0 | — |
case-11 | fail→pass | 9,223 | 8,093 | -12% | 1 | 1 | 0% | 1,610 | 2,955 | +84% | 0 | 0 | — |
case-12 | fail→pass | 9,792 | 9,187 | -6% | 1 | 1 | 0% | 1,765 | 3,195 | +81% | 0 | 0 | — |
case-13 | fail→pass | 11,038 | 3,295 | -70% | 1 | 1 | 0% | 1,680 | 2,106 | +25% | 0 | 0 | — |
case-14 | fail→fail | 10,581 | 2,117 | -80% | 1 | 1 | 0% | 1,810 | 1,806 | -0% | 0 | 0 | — |
case-15 | pass→pass | 12,765 | 2,945 | -77% | 1 | 1 | 0% | 1,992 | 2,048 | +3% | 0 | 0 | — |
case-16 | fail→pass | 14,024 | 11,109 | -21% | 1 | 1 | 0% | 2,592 | 3,214 | +24% | 0 | 0 | — |
case-17 | pass→pass | 5,041 | 4,187 | -17% | 1 | 1 | 0% | 915 | 2,204 | +141% | 0 | 0 | — |
case-18 | pass→pass | 12,080 | 3,303 | -73% | 1 | 1 | 0% | 1,884 | 2,044 | +8% | 0 | 0 | — |
case-19 | pass→pass | 9,495 | 6,129 | -35% | 1 | 1 | 0% | 1,531 | 2,486 | +62% | 0 | 0 | — |
case-20 | fail→pass | 14,945 | 3,111 | -79% | 1 | 1 | 0% | 2,113 | 2,047 | -3% | 0 | 0 | — |
case-21 | fail→pass | 12,994 | 9,602 | -26% | 1 | 1 | 0% | 1,765 | 2,968 | +68% | 0 | 0 | — |
case-22 | fail→pass | 10,886 | 5,707 | -48% | 1 | 1 | 0% | 1,792 | 2,537 | +42% | 0 | 0 | — |
case-23 | pass→pass | 7,781 | 2,246 | -71% | 1 | 1 | 0% | 1,271 | 1,859 | +46% | 0 | 0 | — |
case-24 | fail→pass | 12,863 | 10,326 | -20% | 1 | 1 | 0% | 1,791 | 3,381 | +89% | 0 | 0 | — |
case-25 | pass→pass | 13,347 | 8,551 | -36% | 1 | 1 | 0% | 2,003 | 3,175 | +59% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 25 cases were attempted, and 21 counted toward the lift figure. The other 4 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +40 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.