Menü
Inspect structural blast radius before editing — dependents, clone cohort, risk signals, do-not-touch boundaries. Read-only; does not declare intent.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Basierend auf der SOC-Berufsklassifikation
Human repository setup readiness — status, plan, apply, and wizard over the CLI (not MCP).
Rank demonstrated architectural problems from one canonical run plus targeted source verification. Read-only; priorities are response-local heuristics, never CodeClone findings.
MANDATORY HARD GATE when CodeClone MCP is connected. Before ANY repository file write, obtain edit_allowed=true from start_controlled_change. Before claiming done/verified/ready, finish_controlled_change must accept and clear the intent.
Use CodeClone Engineering Memory through MCP to retrieve scoped evidence before edits and preserve durable incidents, decisions, risks, contradictions, and verification anchors as draft candidates. Memory never grants edit permission; agents propose, humans govern.
Before broad rg/grep, non-trivial planning, unclear scope, contract/schema work, or resuming another agent — call get_implementation_context once to bound the implementation frontier. Then use targeted source reads. Read-only; never grants edit permission.
Maintainer-only — diagnose CodeClone's OWN runtime via Platform Observability after explicit observer enable. Not a user-repo quality signal.
| name | codeclone-blast-radius |
| description | Inspect structural blast radius before editing — dependents, clone cohort, risk signals, do-not-touch boundaries. Read-only; does not declare intent. |
Read-only impact inspection before editing specific files. Does NOT declare intent or start a change workflow.
analyze_repository(root=<abs>) → get_blast_radius(files=[...], depth="transitive")
get_blast_radius is current recomputation from the stored run. When
start_controlled_change returns a slim blast_radius summary with
blast_artifact, use get_blast_artifact(root, run_id, blast_artifact_id) for
the exact full evidence omitted from that start response. Do not substitute a
fresh get_blast_radius call when exact historical start evidence is required.
That slim start response is context_governance.mode="partial_enforce"; the
artifact retrieval response is mode="observe" because it returns the exact
stored object.
For a BUNDLED projection (blast + call_context + memory lanes + freshness + optional
change_control with intent_id), prefer get_implementation_context. Use this skill
when you need ONLY blast fields.
Key / easily-misread fields; the real response carries more.
| Field | Meaning |
|---|---|
radius_level | low / medium / high — overall risk bracket |
origin | the files you asked about |
direct_dependents / transitive_dependents | direct importers / reachable through chains |
clone_cohort_members | shares clone groups with origin — comparison context, NOT edit targets |
in_dependency_cycle | circular-import partners of origin |
structural_risk | high complexity / coupling, low coverage, overloaded modules in the blast zone |
do_not_touch | HARD boundaries — separate explicit approval required |
review_context | supporting context — NOT a ban on editing |
guardrails | actionable review reminders |
blast_artifact | exact retrieval pointer for full start-time blast evidence |
root. No latest run → analyze_repository first.do_not_touch as hard boundaries, clone_cohort_members as comparison, review_context as informational.radius_level high → recommend codeclone-change-control for the actual edit.Radius level + counts → risk signals (if any) → do-not-touch (if any) → guardrails → recommendation (proceed / review dependents first / use change control). Concise.
codeclone-change-control).