MVPavan/via / .claude/skills/execution
execution
UnverifiedUse to execute tracked multi-step work, a standalone plan, or a roadmap phase. Clear bounded implementation can proceed directly under AGENTS.md.
Execution
Execute authorized work through acceptance, verification and recovery. Beads owns durable task state; a plan document is not a prerequisite for a clear task.
Scope selector
| Request | Scope |
|---|---|
Explicit run-all-remaining-phases request or /run-phases | Workstream; read references/workstream-mode.md |
| Execute a named roadmap phase | Phase |
| Task, standalone plan, or single-phase epic | Task |
| Clear bounded implementation without a plan | Task; use the request and acceptance as the packet |
| Material scope/behavior unresolved | Resolve that gap through planning or brainstorming |
Carry authorization through necessary fixes; a review finding cannot authorize
scope expansion. Follow the delegation policy in AGENTS.md.
Risk labels: small is bounded, low-risk work; standard is a bounded behavior
change; deep is cross-cutting, high-risk, or materially unresolved work.
Small work stays inline. Standard work may stay inline or use one bounded worker
when beneficial. Deep work needs risk-appropriate independent review; separate
spec and quality reviewers only when the task requires them or the risks justify
both. Use configured models/effort; if dispatch is unavailable, continue useful
local work and state any required review that remains unperformed.
For delegated units or multi-round review, read references/task-engine.md.
Workspace: scratchpad/execution/<slug>/, with a slug that distinguishes full
plan/workstream paths. Capture the starting revision and dirty-file baseline;
pre-existing edits are not automatically part of the task.
Task loop
- Read the request, existing plan if any, and Beads description/acceptance.
Claim the exact task with
--actor "<runtime>:<session-or-purpose>". - Implement in dependency order, within owned paths. Use test-first or characterization when risk calls for it; apply security to material boundary changes and debugging when a failure's cause is unclear.
- Validate findings as they arrive. Fix confirmed in-scope defects, answer incorrect findings with evidence, and record unresolved decisions. Do not spend fix rounds implementing unverified feedback.
- Run applicable project/plan checks and inspect the scoped final diff. Reuse valid evidence while its code, inputs and environment remain applicable. Small work needs no mandatory reviewer pair. Independent review is driven by explicit requirements or unresolved risk, not the act of closing a task.
- Close with evidence and material limitations in
--reason; report the result. Create durable follow-up work for discoveries outside scope, linked to origin.
For a task epic, map plan tasks to its stage IDs and use the phase close gate.
Phase loop
- Resolve the supplied or unambiguous roadmap and read this phase's deliverables,
spec references, acceptance, risk and exit criterion. Query
bd list -t epic -l ws-<name> --json: exactly one epic title must start with[<phase-id>]. Legacy records may usebd list --spec <roadmap.md> --json. Zero/duplicate matches are a tracking defect; do not silently re-seed. - Confirm the epic is not already closed, check blockers and claim the phase. Elaborate a deep phase when its packet needs a plan; reuse approval covering this work rather than pausing again.
- Select a ready direct child from
bd ready --parent <epic> --json, claim that exact ID, and execute its mapped plan tasks in dependency order. Missing or unmatchedStage:mappings require plan correction before dependent work. - Verify each stage's acceptance before closing it. Regenerate tracking with
the beads skill's
scripts/bd-render-tracking.shwhere available and authorized; report a missing renderer instead of editing generated tracking by hand. - Close gate: query
bd list --parent <epic> --all --json; require at least one stage and every stage closed. Name unclosed stages on failure. Then run the roadmap exit criterion. Close the epic only after both conditions hold, recording the actual exit evidence.
If the governing plan changes or code invalidates the chosen approach, reconcile before dependent work continues. Preserve completed work on failure. An unresolved stage or failed phase exit blocks phase completion and any dependent phase. Commits, publication and broader work retain their existing authorization gates.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 464k | Updated | Claude CodeCodex | Verified | Install |
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 464k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 318k | Updated | Claude CodeCodex | Unverified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 318k | Updated | Claude CodeCodex | Unverified | Install |