CHENG-LIANG1/badass-skills / skills/blast-radius-guard
blast-radius-guard
VerifiedPreflight consequential agent actions by resolving exact targets, authority, reversibility, retry ambiguity, rollback, and independent verification. Use before destructive, externally visible, bulk, financial, permission, deployment, send, or publish operations; do not use for read-only work or trivial reversible local edits.
Blast Radius Guard
Constrain the failure domain before a consequential mutation without slowing ordinary work. The objective is not zero risk; it is a known target set, authorized impact, recoverable execution, and verified outcome.
Decide whether to raise the gate
Evaluate four dimensions:
- Reach: one local artifact, one external object, many objects, or a shared system
- Reversibility: one-step undo, restore from a checkpoint, manual reconstruction, or irreversible
- Ambiguity: deterministic response, delayed result, timeout-prone result, or no reliable status channel
- Authority: explicitly requested target, normal in-scope workflow step, loosely implied mutation, or missing authorization
Use the lowest gate that fits:
- Routine: narrow, reversible, local, and clearly authorized. Proceed with normal verification.
- Guarded: externally visible, multi-object, or costly to reconstruct. Resolve targets from current state and establish an undo path first.
- Hard gate: destructive, financial, permission-changing, broad, irreversible, or authorization-sensitive. Require authority for the exact impact and stop if the target set or recovery story remains unclear.
Do not turn a routine edit into a ceremony. Raise the gate when one dimension is severe or several are uncertain.
Write the mutation contract
Before a guarded or hard-gate action, establish:
- Intent: the outcome the user asked for, not merely the command about to run.
- Exact targets: concrete files, records, recipients, environments, branches, accounts, or resource IDs resolved through read-only inspection.
- Expected delta: what will change and what must remain unchanged.
- Authority: why this mutation is within scope. A broad goal does not authorize a materially different external effect.
- Recovery: undo command, backup, previous revision, provider rollback, or an explicit statement that recovery is impossible.
- Proof: the independent read, remote state, receipt, or behavior that will confirm success.
- Stop trigger: the first unexpected condition that ends the attempt before impact expands.
Keep the contract concise. Never echo secrets, tokens, private payloads, or sensitive recipient data into logs or reports.
Execute with a narrow failure domain
- Prefer exact identifiers over globs, unresolved variables, inferred recipients, or broad directory roots.
- Use a preview, dry run, diff, plan, or read-only API when it accurately models the real action.
- Start with the smallest representative batch when partial rollout is useful and safe.
- Preserve unrelated user state. Do not clean, overwrite, reformat, or restage outside the resolved target set.
- Capture the operation identifier, revision, timestamp, or receipt needed to reconcile the result.
- Stop at the defined trigger. Do not improvise a broader mutation to rescue a failed narrow one.
Handle ambiguous results
A timeout, dropped connection, or missing response does not prove failure. Before retrying a non-idempotent action such as sending, charging, creating, publishing, or appending:
- Query the authoritative current state or operation history.
- Classify the first attempt as applied, not applied, or unknown.
- Retry only when it is confirmed not applied or the provider guarantees idempotency for the same key.
- If state remains unknown, stop and report the reconciliation path instead of risking a duplicate side effect.
For idempotent operations, keep retries bounded and verify that the request key, target, and desired state have not changed.
Verify the outcome independently
Do not treat a successful command exit, redirect, queued status, or optimistic UI as final proof when a source of truth is available.
- Re-read the affected state through an independent path.
- Compare the observed delta with the mutation contract.
- Check the invariants that were supposed to remain untouched.
- For publish or deploy actions, verify the user-facing artifact or remote revision.
- For deletion or replacement, state what disappeared, what replaced it, and whether recovery remains possible.
Report the resolved scope, action taken, verification evidence, and recovery status. If verification fails, say that the mutation was attempted but not confirmed; do not claim completion.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deployment | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |