From grace
Executes GRACE development plans using controller-managed multi-agent parallel waves with selectable safety profiles, verification excerpts, batched artifact sync, and scoped reviews. For parallel module implementation after planning.
npx claudepluginhub osovv/grace-marketplace --plugin graceThis skill uses the workspace's default tool permissions.
Execute a GRACE development plan with multiple agents while keeping planning artifacts and shared context consistent.
Executes GRACE development plans step-by-step with controller-managed context packets, verification excerpts, scoped reviews, level-based verification, and commits after validated steps.
Executes implementation plans at scale via batch with checkpoints, subagent-per-task, parallel dispatch, or TRIZ-parallel modes after skill audits and code context loading.
Orchestrates 4-phase execution loop (IMPLEMENT, VALIDATE, ADVERSARIAL REVIEW, COMMIT) for complex work units with specs. Verifies outputs adversarially in multi-agent setups.
Share bugs, ideas, or general feedback.
Execute a GRACE development plan with multiple agents while keeping planning artifacts and shared context consistent.
docs/development-plan.xml must exist with module contracts and implementation orderdocs/knowledge-graph.xml must existdocs/verification-plan.xml should exist with module-local verification commands and gate expectationsdocs/operational-packets.xml exists, use it as the canonical packet and delta reference$grace-plan themselves before dispatching a large wave$grace-verification themselves before dispatching a large waveParallelize module implementation, not architectural truth.
docs/development-plan.xml, docs/knowledge-graph.xml, docs/verification-plan.xml, phase status, and execution queueIf multiple agents edit the same module, the same shared XML file, or the same tightly coupled slice, this is not a multi-agent wave. Use $grace-execute instead.
Default to balanced unless the user asks otherwise.
safebalanced (default)fastEvery module still gets a fresh worker. Do not optimize this workflow by reusing worker sessions across modules.
Read docs/development-plan.xml, docs/knowledge-graph.xml, and docs/verification-plan.xml once per run, then build the controller view of the execution queue.
Parse pending Phase-N and step-N entries
Group steps into parallel-safe waves
A step is parallel-safe only if:
Choose the execution profile: safe, balanced, or fast
For each wave, prepare a compact execution packet for every module containing:
docs/development-plan.xmldocs/knowledge-graph.xmlDEPENDSdocs/verification-plan.xml, including module-local commands, required scenarios, required log markers, and target test filesWhen docs/operational-packets.xml exists, shape packets and delta proposals to the canonical ExecutionPacket, GraphDelta, and VerificationDelta templates.
Present the proposed waves, selected profile, and packet scopes to the user. In safe, wait for approval before each dispatch. In balanced and fast, one up-front approval is enough unless the plan changes.
Before dispatching, define ownership explicitly:
docs/development-plan.xmldocs/knowledge-graph.xmldocs/verification-plan.xmlIf a worker discovers that a missing module or new dependency is required, stop that worker and ask the user to revise the plan before proceeding. Do not allow silent architectural drift.
For each approved wave:
grace(MODULE_ID): <imperative verb phrase describing what was done>
<File|Function|Export> <name>: <what changed and why>
<File|Function|Export> <name>: <what changed and why>
Every commit body must enumerate the concrete files, functions, or exports that changed and explain what was done to each. Generic phrases like "harden X", "add Y evidence", or "enforce Z" are forbidden. Prefer specific descriptions such as "catch ENOENT in loadPivvConfig and throw CONFIG_NOT_FOUND", "add dedup key collision regression for memory-store.reuseRecord", or "guard write-phase transcript against partial flush".
Keep each checkpoint commit focused on a single coherent change so the history reads like a narrative.Shared-artifact rule for workers:
After each worker finishes:
$grace-reviewer audit only when:
After all modules in the wave are approved:
docs/knowledge-graph.xmldocs/verification-plan.xmldocs/development-plan.xml step status once per wave$grace-refresh against the changed modules and touched dependency surfaces$grace-verification themselves before continuingRun verification at the smallest level that still protects correctness.
Do not run full-repository tests and full-repository graph scans after every successful module unless the risk profile requires it.
After each wave, the controller commits only shared artifacts that changed:
docs/knowledge-graph.xml, docs/verification-plan.xml, and docs/development-plan.xml with wave resultsgrace(meta): sync <artifacts updated> after wave N
<module-id>: <what changed in the graph/plan/verification for this module>
<module-id>: <what changed in the graph/plan/verification for this module>
List only the modules whose shared-artifact entries actually changed and describe what was updated (e.g. "M-CONFIG: marked step-1 complete, added public exports to graph"). Omit modules with no delta.Worker implementation commits are already done per module in Step 3. Controller commits are only for shared planning artifacts.
After each wave, report:
=== WAVE COMPLETE ===
Wave: N
Profile: safe / balanced / fast
Modules: M-xxx, M-yyy
Approved: count/count
Graph sync: targeted passed / targeted fixed / escalated to full refresh
Verification: module-local passed / wave checks passed / follow-up required
Remaining waves: count
safe rather than pretending fast is safeUse $grace-execute for sequential execution when dependency risk is higher than the parallelism gain.