Sweeps architecture for impact before a change — spawns agents per C3 entity for constraint-checked analysis. Advisory only — does NOT make changes (route to c3-change for implementation). This skill should be used when the user asks to: - "assess C3 impact of X", "what would break in C3 if..." - "check C3 constraints for X", "is this change safe per C3 docs" - "affected C3 components", "which C3 containers are impacted by X" - "sweep C3 architecture", "impact assessment against C3 architecture" <example> user: "assess C3 impact if I replace the auth system" assistant: "Using c3-sweep to assess architectural impact." </example> DO NOT use for: "what is X" / "explain X" (c3-query), implementation (c3-change), patterns (c3-ref). Requires .c3/ to exist.
Install
npx skillscat add lagz0ne/c3-skill/c3-sweep Install via the SkillsCat registry.
C3 Sweep: Impact Assessment Team
Sweep the architecture for impact of proposed changes through a team of agents that read C3 docs dynamically.
Precondition: C3 Adopted
Run npx -y @lagz0ne/c3x list --json via Bash. STOP if it fails or returns empty.
If missing:
This project doesn't have C3 architecture docs yet. Use the c3-onboard skill to create documentation first.
How It Works
This skill creates an Agent Team with a lead and specialized workers:
You (user)
↕ conversation
Team Lead (c3-sweep-lead agent, delegate mode)
↕ coordinates
Workers:
- Container workers (read container docs, delegate to component workers)
- Component workers (inspect code against docs + refs)
- Ref workers (check convention compliance)Execution
HARD RULE: Your FIRST action must be to spawn the c3-sweep-lead agent. Do not read C3 docs yourself, do not create teams yourself, do not assess impact yourself. The lead handles everything.
Spawn the lead:
Task tool:
subagent_type: "c3-skill:c3-sweep-lead"
prompt: "<pass the user's full change request / question here>"
mode: "delegate"The lead will:
- Load topology — Run
npx -y @lagz0ne/c3x list --jsonto get full system structure (entities, relationships, frontmatter) - Identify entities — Match the change to affected containers, refs, and ADRs from the JSON
- Delegate — Spawn container + ref workers in parallel for deep inspection
- Synthesize — Collect advisories into a unified impact assessment
The lead communicates back when it needs clarification. Relay these to the user and pass their responses back.
Team Configuration
The lead operates in delegate mode (coordination only, never modifies code or docs). The lead creates its own Agent Team via TeamCreate and spawns workers into it. Workers read C3 docs directly.
Worker Tiers
| Worker | Role |
|---|---|
| Container | Reads container README, identifies affected components, spawns component workers |
| Component | Inspects actual code against component doc + applicable refs |
| Ref | Checks whether the change complies with or violates conventions |
Assessment Output
The lead synthesizes all worker advisories into:
- Affected Entities — which containers and components, with reasons
- Constraint Chain — all conventions, refs, and ADRs that apply
- File Changes — specific files that would need modification
- Risks — edge cases, relationship impacts, ADR conflicts
- Recommended Approach — step-by-step plan respecting all constraints
Routing
- Want to implement after assessment? → c3-change skill
- Architecture questions without change context → c3-query skill
- Pattern management → c3-ref skill
- Standalone audit → c3-audit skill