sprngr/rubber-duck / skills/quack
quack
VerifiedExplicit user-invoked routing for Rubber Duck workflows. Use when: "quack", "quack <intent>".
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.1 KB |
| assets/heartbeat.md | markdown | 1.5 KB |
| assets/quick-help.md | markdown | 1.4 KB |
| assets/route-aliases.json | config | 2.3 KB |
| assets/subagent-runbook.md | markdown | 5.7 KB |
| references/Examples.md | markdown | 2.6 KB |
| references/GUARDRAILS.md | markdown | 757 B |
assets/subagent-runbook.md
View on GitHub# Subagent Runbook (quack inline-injection)
Purpose: role instructions appended inline by `quack` when invoking a subagent.
Usage rule:
- `quack` resolves `route`, then determines effective `preferred_subagent` (user override or default `duckling`).
- `quack` selects the matching role section below.
- `quack` appends role instructions inline to subagent invocation payload.
Return handling:
- After subagent completes, control returns to the parent agent. Return-side handling lives in `rubber-duck` body under "Subagent Return Handling". `quack`'s Method has no return-side responsibilities.
- If parent is not `rubber-duck`, host project defines its own return-side flow.
Role coverage:
- active skills only: `duck-patch`, `duck-debug`, `duck-design`, `duck-risk`, `duck-simplify`, `duck-teach`, `duck-triage`, `duck-debt`, `duck-review`
---
## role: patch
Intent:
- Execute smallest safe implementation patch in bounded scope.
Constraints:
- phase-gated scope with adaptive caps:
- Phase 1 (stubs/interfaces): up to 6 files
- Phase 2 (wiring/integration): up to 4 files
- Phase 3 (concrete implementation): up to 2 files
- no scope expansion without renewed approval
- preserve security/trust/data-loss/accessibility requirements
Output shape:
- changed files
- behavior delta
- smallest verification result
- blocker question if blocked
Handoff cues:
- if root cause unclear, request `duck-debug` trace-mode evidence first.
---
## role: debug
Intent:
- Run ask-first root-cause debugging with evidence-first progression.
Constraints:
- questions before fix direction when context is incomplete
- no patch recommendation until evidence/call-path is clear
- when asked for trace-only, enforce read-only facts mode
Output shape:
- 1-3 targeted questions
- likely execution path to inspect
- one falsifiable next check
- optional one-line root-cause statement when identified
Handoff cues:
- if user requests implementation, route to `patch` after bounded approval.
---
## role: design
Intent:
- Evaluate approach tradeoffs and constraints before implementation.
Constraints:
- no hidden architecture decisions
- keep options evidence-backed with explicit tradeoffs
Output shape:
- compact option matrix (pros/cons/risks)
- recommended next decision question
Handoff cues:
- route chosen implementation to `patch`; route risk stress-test to `risk`.
---
## role: risk
Intent:
- Adversarially stress-test failure modes, rollback, compatibility, and trust boundaries.
Constraints:
- max 3 highest-impact findings
- no style/naming nits
Output shape:
- findings with `Impact`, `Rollback`, and `Fix`
- trust-boundary/rollback coverage line
Handoff cues:
- route implementation to `patch` after mitigation direction is accepted.
---
## role: simplify
Intent:
- Reduce overengineering and complexity tax safely.
Constraints:
- do not weaken security/correctness/explicit requirements
- prefer deletion/reuse/stdlib/native before new abstraction
Output shape:
- simplification findings with smallest-shape fix
- estimated net reduction line
Handoff cues:
- if duplication-specific, stay in `simplify` dry mode.
---
## role: teach
Intent:
- Teach or explain with depth-scaled structure.
Constraints:
- keep sectioned format explicit
- prefer workspace patterns over generic examples
- no edits/actions in teaching flow
Output shape:
- explain mode: `What / Why / Watch out / Next question`
- show/teach mode: `What / Why / Example / Pitfalls / See also`
Handoff cues:
- if explanation uncovers breakage, route to `debug`.
---
## role: triage
Intent:
- Classify severity and identify highest-value missing tests.
Constraints:
- prioritize correctness/security/data-loss regressions first
- no patching; test-planning and prioritization only
Output shape:
- severity classification + rationale
- missing test scenarios
- one minimum runnable check suggestion
Handoff cues:
- route implementation of agreed fix to `patch`
- route final changed-code comments to `review`
---
## role: debt
Intent:
- Build read-only deferred-work ledger from TODO/FIXME/HACK/XXX markers.
Constraints:
- read-only audit; no prioritization disguised as implementation mandate
- strict mode only returns issue-linked entries when requested
Output shape:
- grouped debt entries with location and marker type
- coverage note (scanned scope / strict vs broad)
Handoff cues:
- route concrete fix request from debt item to `patch` or `review`.
---
## role: review
Intent:
- Consolidate final changed-code review findings.
Constraints:
- changed-code scope only
- strongest-risk-first deduped comments
Output shape:
- schema-first one-line findings: prefix + location + problem + `Fix:`
- Auto-Clarity only for security/irreversible-risk detail
Handoff cues:
- route accepted implementation work to `patch`.
---
## role: refactor
Intent:
- Execute bounded code restructuring with reference tracking.
Refactoring operations:
- extract function/class/method
- rename across codebase
- move code between files (with import updates)
- inline function/variable
- convert patterns (callback -> promise, class -> hooks)
Constraints:
- trace all references before refactoring (like debug trace mode)
- phase-gated scope with adaptive caps (6/4/2 by phase); split when phase caps or review-fatigue triggers are exceeded
- require execution approval with full change plan
- distinguish from single-file patch and complexity reduction
Output shape:
- refactoring plan: operation, target, destination, files affected, verification
- reference trace: N references across M files
- verification report: tests passed, imports correct
Handoff cues:
- if refactoring requires design decisions -> route to `design` first
- if refactoring reveals complexity issues -> note for `simplify` follow-up
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |