AgentStack
SKILL verified MIT Self-run

Composing Siblings

skill-panqiwei-board-superpowers-composing-siblings · by PanQiWei

|

No reviews yet
0 installs
11 views
0.0% view→install

Install

$ agentstack add skill-panqiwei-board-superpowers-composing-siblings

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Composing Siblings? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

composing-siblings

This skill is the single source of truth for invoking sibling-plugin skills (gstack:* / superpowers:*) from any board-superpowers context. It does not perform actions itself; it provides the invocation contract that callers follow before routing to a sibling.

How to apply this skill

| What you need | Where to look | |---------------|---------------| | Which sibling skill to invoke at a specific handoff | references/handoff-points.md (9 caller × scenario table) | | Current gstack:* / superpowers:* skill names and descriptions | references/sibling-plugin-table.md | | Whether a sibling skill is safe to invoke from Mode-2 Consumer | references/procedural-fallback-rules.md | | Namespace prefix rules + atomic reflex constraint | references/boundary.md |

Invocation rules

Rule 1 — SKILL invocation, not subagent spawn. All cross-plugin composition uses SKILL invocation. The platform's skill matcher loads the sibling SKILL.md body into the calling agent's context as procedural guidance — no second context window, no Agent tool call. This is content- loading, not a spawn. The max_depth budget applies only to Agent tool calls; SKILL invocation does not consume depth.

Rule 2 — Always carry the namespace prefix. Every sibling-plugin reference uses : form:

  • superpowers:test-driven-development (correct)
  • gstack:/review (correct — gstack uses / prefix by convention)
  • test-driven-development (wrong — bare reference fails cross-plugin resolution)

Rule 3 — Procedural check before Mode-2 invocation. A Mode-2 Consumer runs as a CC subagent (max_depth=1). SKILL invocation itself does not consume depth, but any spawn-instruction inside the invoked skill's body would. Before invoking any sibling from a Mode-2 context, verify it is procedural (its body does not instruct an Agent / spawn_agents_on_csv call). Consult references/procedural-fallback-rules.md for the current status of each sibling.

Rule 4 — Phase-based routing. Route by phase, not by preference:

  • Bookend phases (direction-setting before a card is claimed; delivery-side

QA / review / security) → gstack:/* skills.

  • Middle phase (implementation loop: brainstorming → writing-plans → TDD →

debugging → verification → code-review) → superpowers:* skills.

  • Conflict arbitration: user instructions > skill > default behavior. A gstack

skill's "start coding" output does not override superpowers:test-driven-development discipline unless the user explicitly says so in the current conversation.

Handoff points quick reference

All current callers in the v0.7.0 Producer/Consumer surfaces use this skill. See references/handoff-points.md for the full table. High-frequency summary:

| Caller | Phase label | Primary sibling(s) | |--------|-------------|-------------------| | intaking-requirement (intake) | B1 — direction question | gstack:/office-hours, gstack:/plan-ceo-review | | intaking-requirement (intake) | B1 — architecture question | gstack:/plan-eng-review | | intaking-requirement (intake) | B1 — plan synthesis | superpowers:brainstorming, superpowers:writing-plans | | consuming-card | C1 — planning | superpowers:writing-plans | | consuming-card | C2 — implementation | superpowers:subagent-driven-development, superpowers:test-driven-development | | consuming-card | C3 — pre-PR verification | superpowers:verification-before-completion, superpowers:requesting-code-review, gstack:/review | | consuming-card | C4 — conditional QA/security | gstack:/qa (UI cards), gstack:/cso (security-flagged) | | decomposing-into-milestones | plan synthesis | superpowers:writing-plans | | decomposing-into-milestones | arch validation | gstack:/plan-eng-review |

What this skill does NOT cover

  • Which sibling skill wins when two could apply — that's the caller's

routing logic using the phase-based rule above.

  • Whether to use a sibling at all — that's the calling skill's decision

based on the card's Execution Hints and the architect's signal.

  • Autonomy classification of sibling-skill invocations — that's

board-superpowers:classifying-actions.

  • Audit rows for sibling-skill handoffs — that's

board-superpowers:auditing-actions.

  • Same-plugin (board-superpowers internal) skill references — those follow

the SKILLS.md call-graph topology, not this skill.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.