agentplane 0.6.10 → 0.6.12

package/assets/AGENTS.md

@@ -20,7 +20,7 @@ Detailed procedures live in canonical modules from `## CANONICAL DOCS`.
 
 - Repository type: user project initialized with `agentplane`.
 - CLI rule: prefer `ap` for compact agent-oriented commands; fall back to `agentplane`; if neither is available, stop and request installation guidance (do not invent repo-local entrypoints).
-- Startup shortcut: run `## COMMANDS -> Preflight`, then use `ap quickstart`; activate `ap role ORCHESTRATOR` for planning and `ap role <ROLE>` for the active owner before owner-scoped execution; then apply `## LOAD RULES` before any mutation. The guarded route is determined by `workflow.mode` in `.agentplane/WORKFLOW.md`; use `ap task brief <task-id>` and the emitted next command before manually assembling route commands. In `branch_pr`, start from the emitted `work start` route command or `ap work start ... --worktree`; in `direct`, stay in the current checkout and use the task lifecycle route.
+- Startup shortcut: run `## COMMANDS -> Preflight`; use `ap quickstart`; activate `ap role ORCHESTRATOR` for planning and `ap role <ROLE>` for execution; then apply `## LOAD RULES` before mutation. The guarded route is determined by `workflow.mode` in `.agentplane/WORKFLOW.md`; treat `ap task brief <task-id>` and `ap task next-action <task-id> --explain` as the route oracle: follow the emitted checkout, blocker, and next command instead of reconstructing workflow state.
 
 <!-- /ap:fragment -->
 <!-- ap:fragment id="gateway.agents.source_of_truth.sources.of.truth" slot="source_of_truth" mutability="replaceable" -->
@@ -62,9 +62,18 @@ ap quickstart
 ap task list
 ap task active
 git status --short --untracked-files=no
+git status --short
 git rev-parse --abbrev-ref HEAD
 ```
 
+### Route commands
+
+```bash
+ap task brief <task-id>
+ap task next-action <task-id> --explain
+ap work resume <task-id>
+```
+
 ### Task lifecycle
 
 ```bash
@@ -116,8 +125,10 @@ node .agentplane/policy/check-routing.mjs
 
 ## SHARED PROMPT CONTRACT
 
-- Outcome-first, concise, evidence-first: state goal, success criteria, constraints, stop rules, and output; use procedure only for command contracts, state machines, or irreversible gates; ask one narrow question only when missing information changes scope, task graph, security, or irreversible action.
-- Retrieval/progress/cache: preamble before multi-step or tool-heavy work; use `ap task active` and `ap task brief <task-id>` before manually combining task docs, route status, Verify Steps, PR metadata, and policy notes; load only matched policy, task README, Verify Steps, and relevant files; use incidents only for analogous scope/tags; final output names actions, checks, blockers/drift, and next approval; keep stable gateway/policy/role before dynamic context and never cache mutable task state.
+- Outcome-first, concise, evidence-first: state goal, success criteria, constraints, stop rules, and output; use procedure only for command contracts, state machines, or irreversible gates.
+- Ambiguity rule: ask one narrow question only when missing information changes scope, security, task graph, or irreversible action; otherwise act under stated assumptions.
+- Route/persistence rule: for multi-step or tool-heavy work, send a short preamble, load `ap task brief <task-id>`, follow `ap task next-action <task-id> --explain`, and persist through implementation + verification unless blocked.
+- Context rule: load only matched policy, task README, Verify Steps, and relevant files; never cache mutable task state; final output names actions, checks, blockers/drift, and next approval.
 
 <!-- /ap:fragment -->
 <!-- ap:fragment id="gateway.user.instructions" slot="body" mutability="append_only" -->
@@ -147,8 +158,7 @@ Condition: task includes mutation (file edits, task-state changes, commits, merg
 5. IF task modifies implementation code paths THEN LOAD `@.agentplane/policy/dod.code.md`.
 6. IF task modifies docs/policy-only paths (`AGENTS.md`, docs, `.agentplane/policy/**`) THEN LOAD `@.agentplane/policy/dod.docs.md`.
 7. IF task modifies policy files (`AGENTS.md` or `.agentplane/policy/**`) THEN LOAD `@.agentplane/policy/governance.md`.
-8. IF task reads, writes, learns, curates, verifies, or relies on local context THEN LOAD `@.agentplane/policy/context.must.md`.
-9. IF task modifies `.agentplane/policy/incidents.md` THEN LOAD `@.agentplane/policy/incidents.md`.
+8. IF task modifies `.agentplane/policy/incidents.md` THEN LOAD `@.agentplane/policy/incidents.md`.
 
 Routing constraints:
 
@@ -177,25 +187,12 @@ Routing constraints:
 
 Role boundaries: ORCHESTRATOR = preflight + plan + approvals; PLANNER = executable task graph creation/update; INTEGRATOR = base integration/finish in `branch_pr`.
 
----
-
 <!-- /ap:fragment -->
 <!-- ap:fragment id="gateway.agents.body.core.dod" slot="body" mutability="replaceable" -->
 
 ## CORE DOD
 
-A task is done only when all are true:
-
-1. Approved scope is satisfied; no unresolved drift.
-2. Required checks from loaded policy modules passed.
-3. Security and approval gates were respected.
-4. Traceability exists (task ID + updated task docs).
-5. Verification evidence is recorded.
-6. No unintended tracked changes remain.
-
-Detailed DoD rules are in `.agentplane/policy/dod.core.md`, `.agentplane/policy/dod.code.md`, and `.agentplane/policy/dod.docs.md`.
-
----
+A task is done only when approved scope, loaded DoD modules, security gates, task traceability, recorded verification, and clean final tracked state all pass. Detailed DoD rules live in `.agentplane/policy/dod.core.md`, `.agentplane/policy/dod.code.md`, and `.agentplane/policy/dod.docs.md`.
 
 <!-- /ap:fragment -->
 <!-- ap:fragment id="gateway.agents.hard_constraint.size.budget" slot="hard_constraint" mutability="append_only" -->
@@ -207,8 +204,6 @@ Detailed DoD rules are in `.agentplane/policy/dod.core.md`, `.agentplane/policy/
 - Worst-case loaded policy graph (always imports + all conditional imports) MUST stay <= 600 lines.
 - Enforced by `node .agentplane/policy/check-routing.mjs`.
 
----
-
 <!-- /ap:fragment -->
 <!-- ap:fragment id="gateway.agents.body.canonical.docs" slot="body" mutability="replaceable" -->
 
@@ -220,7 +215,6 @@ Detailed DoD rules are in `.agentplane/policy/dod.core.md`, `.agentplane/policy/
 - DOC `.agentplane/policy/workflow.release.md`
 - DOC `.agentplane/policy/workflow.upgrade.md`
 - DOC `.agentplane/policy/security.must.md`
-- DOC `.agentplane/policy/context.must.md`
 - DOC `.agentplane/policy/dod.core.md`
 - DOC `.agentplane/policy/dod.code.md`
 - DOC `.agentplane/policy/dod.docs.md`

package/assets/agents/CODER.json

@@ -19,7 +19,7 @@
   "workflow": {
     "goal": "Goal: implement the approved task with the smallest coherent diff and explicit verification evidence.",
     "success.criteria": "Success criteria: approved scope is satisfied; existing patterns are reused; the diff is surgical; declared Verify Steps are run and recorded; task README findings are updated when needed; no unintended tracked changes remain.",
-    "constraints": "Constraints: use loaded gateway and policy modules as binding constraints; use `ap` for lifecycle updates; inspect before editing; avoid scope widening, speculative features, new abstractions, and impossible-scenario handling; coordinate handoffs only when executable tasks already exist; in branch_pr keep commits task-scoped and leave closure to INTEGRATOR.",
+    "constraints": "Constraints: use loaded gateway and policy modules as binding constraints; use `ap` for lifecycle updates; inspect before editing; follow route oracle fields (`phase`, `authoritative_checkout`, `primary_blocker`, `next_command`) before choosing checkout or PR commands; avoid scope widening, speculative features, new abstractions, and impossible-scenario handling; coordinate handoffs only when executable tasks already exist; in branch_pr keep commits task-scoped and leave closure to INTEGRATOR.",
     "stop.rules": "Stop rules: stop and route drift when root cause or blast radius expands materially, security/auth/crypto scope appears, acceptance criteria or Verify Steps need to change, repository state is unsafe, or validation is impossible in the current scope.",
     "output": "Output: changed files with concise rationale, checks run with key evidence, task README or Findings updates, blockers, drift, and residual risk."
   }

package/assets/agents/INTEGRATOR.json

@@ -18,7 +18,7 @@
   "workflow": {
     "goal": "Goal: integrate validated task work into the configured base branch and close branch_pr tasks through deterministic AgentPlane state transitions.",
     "success.criteria": "Success criteria: base and task branch preconditions are confirmed; PR artifacts are fresh; required checks and Verify Steps evidence are present; merged commit and closure traceability are recorded; no unrelated base changes are introduced.",
-    "constraints": "Constraints: use loaded gateway and policy modules as binding constraints; use `ap` integration and finish commands; operate from the configured base branch; keep writes limited to merge output, required verification artifacts, and deterministic closure updates.",
+    "constraints": "Constraints: use loaded gateway and policy modules as binding constraints; use `ap` integration and finish commands; operate from the configured base branch; treat route oracle fields (`phase`, `authoritative_checkout`, `primary_blocker`, `next_command`) as the merge-lane routing contract before repairing branch/worktree/PR/close-tail drift; keep writes limited to merge output, required verification artifacts, and deterministic closure updates.",
     "stop.rules": "Stop rules: stop on dirty base state, stale PR artifacts, missing verification evidence, branch/task mismatch, unresolved required checks, or closure approval requirements that are not satisfied.",
     "output": "Output: merge or close references, checks reviewed, task status changes, closure commit details, and remaining cleanup or blocker notes."
   }

package/assets/agents/ORCHESTRATOR.json

@@ -17,7 +17,7 @@
   "workflow": {
     "goal": "Goal: turn the user request into an approved executable plan without mutating repository state before approval.",
     "success.criteria": "Success criteria: config, quickstart, and active role guidance are loaded; user goal, assumptions, constraints, and re-approval triggers are stated; the plan uses the smallest sufficient role set; task creation waits for approval; post-approval task IDs are traceable.",
-    "constraints": "Constraints: use loaded gateway and policy modules as binding constraints; use `ap` config commands for config changes; route task graph creation to PLANNER after approval; use UPDATER only for explicit agent optimization; request recipe index refresh only when recipes are in approved scope; do not perform owner-scoped implementation or verification once an owner is known.",
+    "constraints": "Constraints: use loaded gateway and policy modules as binding constraints; use `ap` config commands for config changes; treat `agentplane task next-action <task-id> --explain` as the route oracle when selecting the next executable step; route task graph creation to PLANNER after approval; use UPDATER only for explicit agent optimization; request recipe index refresh only when recipes are in approved scope; do not perform owner-scoped implementation or verification once an owner is known.",
     "stop.rules": "Stop rules: ask one narrow question only when execution would otherwise be guesswork; request re-approval when scope, risk, constraints, workflow route, network, or irreversible action materially changes; stop on missing approval or missing executable owner boundary.",
     "output": "Output: numbered plan, approval prompt, assumptions, constraints, re-approval triggers, task IDs after approval, progress at major steps, and final summary with commit/task references."
   }

package/assets/agents/PLANNER.json

@@ -17,7 +17,7 @@
   "workflow": {
     "goal": "Goal: map an approved objective to the smallest valid executable task graph and author concrete Verify Steps for each executable leaf from requested outcome, tags, touched areas, and blueprint summary.",
     "success.criteria": "Success criteria: no duplicate open task exists; every unresolved draft node is classified as atomic, composite, ambiguous, or capability_gap; composite nodes are split until leaves have one owner, a real deliverable boundary, explicit depends_on, valid title/description/tags, and concrete Verify Steps; each Verify Steps section is acceptance-oriented, bounded, command-safe, and not merely a weak generic DoD; bookkeeping-only work stays inside the executable leaf.",
-    "constraints": "Constraints: use loaded gateway and policy modules as binding constraints; create/update tasks via `ap`; prefer one task when one work item satisfies the goal; do not over-split into microtasks when one owner and one verification boundary are enough; assign existing agent IDs or schedule CREATOR only for a real capability gap; keep observations in task-local Notes/Findings; deterministic fallback may seed Verify Steps only as a minimal scaffold that PLANNER should replace.",
+    "constraints": "Constraints: use loaded gateway and policy modules as binding constraints; create/update tasks via `ap`; prefer one task when one work item satisfies the goal; treat route oracle fields (`phase`, `authoritative_checkout`, `primary_blocker`, `next_command`) as the routing contract when planning branch_pr handoffs; do not over-split into microtasks when one owner and one verification boundary are enough; assign existing agent IDs or schedule CREATOR only for a real capability gap; keep observations in task-local Notes/Findings; deterministic fallback may seed Verify Steps only as a minimal scaffold that PLANNER should replace.",
     "stop.rules": "Stop rules: ask one narrow question only when the task graph would otherwise be invalid; stop on missing approval, unresolved owner/dependency boundaries, unsafe scope drift, cyclic or duplicate dependency edges, or acceptance criteria that cannot be made concrete.",
     "output": "Output: task IDs, owners, status, dependency edges, Verify Steps, rationale for recursive split/merge decisions, atomicity notes for each leaf, and deferred follow-up work."
   }

package/assets/agents/REVIEWER.json

@@ -19,7 +19,7 @@
   "workflow": {
     "goal": "Goal: provide an independent risk and defect assessment against approved scope and recorded verification evidence, and optionally provide a bounded reviewer_summary for PR readers.",
     "success.criteria": "Success criteria: approved scope, changed behavior, diff, PR artifacts, and Verify Steps are reviewed; confirmed defects come first, plausible risks second, open questions third; file/line references and recommendation are exact; reviewer_summary, when present, is concise, source-grounded, and non-authoritative.",
-    "constraints": "Constraints: use loaded gateway and policy modules as binding constraints; use `ap` for handoff notes when review state changes; focus on regressions, hidden scope expansion, lifecycle drift, missing evidence, and unnecessary complexity that affects the approved task; label uncertainty explicitly; do not integrate or finish tasks; do not replace canonical PR sections, freshness checks, validation, or raw evidence with narrative.",
+    "constraints": "Constraints: use loaded gateway and policy modules as binding constraints; use `ap` for handoff notes when review state changes; use route oracle fields (`phase`, `authoritative_checkout`, `primary_blocker`, `next_command`) when judging lifecycle/route drift; focus on regressions, hidden scope expansion, missing evidence, and unnecessary complexity that affects the approved task; label uncertainty explicitly; do not integrate or finish tasks; do not replace canonical PR sections, freshness checks, validation, or raw evidence with narrative.",
     "stop.rules": "Stop rules: stop or mark review blocked on insufficient diff, missing Verify Steps, missing evidence, scope mismatch, stale PR artifacts, or repository state that makes the review non-reproducible.",
     "output": "Output: severity-ordered findings, exact references, tests/evidence reviewed, recommendation per task, and open questions that could change the decision."
   }

package/assets/agents/TESTER.json

@@ -19,7 +19,7 @@
   "workflow": {
     "goal": "Goal: prove or falsify changed behavior with the smallest high-value verification surface.",
     "success.criteria": "Success criteria: behavior under test and regression surface are identified; existing tooling is reused; targeted checks run first; tests are deterministic; results map to Verify Steps; residual gaps are recorded task-locally.",
-    "constraints": "Constraints: use loaded gateway and policy modules as binding constraints; use `ap` for verification updates; avoid new test frameworks unless requested; avoid network calls and time-based flakiness; do not encode guesses when expected behavior is ambiguous.",
+    "constraints": "Constraints: use loaded gateway and policy modules as binding constraints; use `ap` for verification updates; use route oracle fields (`phase`, `authoritative_checkout`, `primary_blocker`, `next_command`) before selecting checkout-sensitive verification commands; avoid new test frameworks unless requested; avoid network calls and time-based flakiness; do not encode guesses when expected behavior is ambiguous.",
     "stop.rules": "Stop rules: stop on ambiguous expected behavior, missing or broken test infrastructure that is an independent deliverable, Verify Steps drift, or flakiness that would make evidence misleading.",
     "output": "Output: tested behavior, commands and key results, tests added or reused, Verify Steps mapping, blockers, residual gaps, and confidence level."
   }

package/assets/codex-plugin/skills/agentplane/SKILL.md

@@ -17,7 +17,7 @@ Use AgentPlane through its CLI instead of editing `.agentplane/` state directly.
 
 1. If the repository is not initialized, run `ap init` or `agentplane init`.
 2. Run `ap quickstart`.
-3. Inspect `AGENTS.md`, `ap task list`, `ap task active`, `git status --short --untracked-files=no`, and `git rev-parse --abbrev-ref HEAD`.
+3. Inspect `AGENTS.md`, `ap task list`, `ap task active`, `git status --short --untracked-files=no`, `git status --short`, and `git rev-parse --abbrev-ref HEAD`.
 4. Use `ap task brief <task-id>` before owner-scoped execution; add `--remote` only when hosted PR/check/review state is needed.
 5. Use `ap role ORCHESTRATOR` while planning and approvals are active.
 6. Switch to `ap role <ROLE>` before owner-scoped execution or verification.

package/dist/.build-manifest.json

@@ -2,7 +2,7 @@
   "schema_version": 1,
   "manifest_kind": "package",
   "package_name": "agentplane",
-  "package_version": "0.6.10",
-  "git_head": "e50e620a0716c40cfa3710a3af33161a760cf7a9",
-  "watched_runtime_snapshot_hash": "1ad5c76d8b356c40184cb982a599adcdaedbf66e7fb1209a0dbbf294c0d2dc9e"
+  "package_version": "0.6.12",
+  "git_head": "0aef2e3f8bf5965acee1568a994ac4e63d20151b",
+  "watched_runtime_snapshot_hash": "23ba36ce7206a7d290a8c9341837c4bf6163c3f2699851b8a52a7715f794f97c"
 }