opencode-plugin-flow 4.1.12 → 4.1.14

package/CHANGELOG.md

@@ -3,6 +3,14 @@
 One short line per release. For the full rationale behind each entry, see the
 commit history and review evidence.
 
+## [4.1.14] - 2026-06-22
+
+Publish Flow through npm trusted publishing so release CI no longer depends on expiring npm tokens.
+
+## [4.1.13] - 2026-06-21
+
+Teach Flow's parallel orchestration lore to walk a full wave while pinning hidden worker permissions to tested docs.
+
 ## [4.1.12] - 2026-06-18
 
 Harden Flow's hidden worker lore, scriptable doctor checks, typed package smoke, and session edge contracts without widening the v4 runtime surface.

package/README.md

@@ -21,8 +21,8 @@ The manager still owns every Flow state change. Workers gather evidence; they do
 Use OpenCode's plugin installer when your OpenCode version supports it:
 
 ```bash
-opencode plugin opencode-plugin-flow@4.1.12 --global --force
-npx -y opencode-plugin-flow@4.1.12 sync
+opencode plugin opencode-plugin-flow@4.1.14 --global --force
+npx -y opencode-plugin-flow@4.1.14 sync
 ```
 
 The first command adds Flow to your global OpenCode plugin config or replaces an
@@ -48,7 +48,7 @@ OpenCode config manually instead:
 
 ```json
 {
-  "plugin": ["opencode-plugin-flow@4.1.12"]
+  "plugin": ["opencode-plugin-flow@4.1.14"]
 }
 ```
 
@@ -59,7 +59,7 @@ duplicate entry.
 Then run the same pre-start skill sync and start or restart OpenCode:
 
 ```bash
-npx -y opencode-plugin-flow@4.1.12 sync
+npx -y opencode-plugin-flow@4.1.14 sync
 ```
 
 Project-local skill overrides still work through OpenCode's normal lookup:
@@ -84,15 +84,15 @@ version in your global `opencode.json`.
 To inspect the installed skill set:
 
 ```bash
-npx -y opencode-plugin-flow@4.1.12 doctor
+npx -y opencode-plugin-flow@4.1.14 doctor
 ```
 
 For automation, keep the default human-readable `doctor` output and opt into
 machine behavior explicitly:
 
 ```bash
-npx -y opencode-plugin-flow@4.1.12 doctor --json
-npx -y opencode-plugin-flow@4.1.12 doctor --check
+npx -y opencode-plugin-flow@4.1.14 doctor --json
+npx -y opencode-plugin-flow@4.1.14 doctor --check
 ```
 
 `doctor --check` and `doctor --strict` exit nonzero when the health status is
@@ -111,7 +111,7 @@ above first. Missing, incomplete, or outdated managed skills can still be
 repaired with:
 
 ```bash
-npx -y opencode-plugin-flow@4.1.12 sync
+npx -y opencode-plugin-flow@4.1.14 sync
 ```
 
 Then restart OpenCode so the refreshed registry is loaded. `sync` manages all
@@ -226,7 +226,7 @@ First remove `opencode-plugin-flow` from your OpenCode plugin config so future
 OpenCode startups stop loading Flow. Then remove Flow-owned synced skills:
 
 ```bash
-npx -y opencode-plugin-flow@4.1.12 uninstall
+npx -y opencode-plugin-flow@4.1.14 uninstall
 ```
 
 Restart OpenCode after both steps. This removes Flow-owned synced skills when

package/dist/cli.js

@@ -1,11 +1,20 @@
 #!/usr/bin/env node
-import{createHash as De}from"node:crypto";import{mkdir as Te,readdir as oe,readFile as Pe,rm as je,writeFile as b}from"node:fs/promises";import{createRequire as Se}from"node:module";import{dirname as We,join as u,normalize as ze,sep as Be}from"node:path";var S=`# Flow worker handoff contract
+import{createHash as De}from"node:crypto";import{mkdir as Te,readdir as oe,readFile as Pe,rm as je,writeFile as b}from"node:fs/promises";import{createRequire as Se}from"node:module";import{dirname as We,join as u,normalize as Oe,sep as Be}from"node:path";var S=`# Flow worker handoff contract
 
 Flow managers merge only the worker's final response. Treat that response as the
 worker report of record: it must include the assigned scope, what was actually
 covered, the evidence for each useful claim, and the remaining gaps. End worker
 prompts with "Return only this Flow handoff."
 
+Status meanings:
+
+- \`success\`: the assigned scope was covered, or any skipped items are explicitly
+  immaterial to the assigned question.
+- \`partial\`: useful evidence was gathered, but material assigned scope remains
+  unchecked or unresolved.
+- \`blocked\`: the worker cannot answer the assigned question without missing
+  access, input, dependencies, or manager clarification.
+
 ## Evidence, review, validation, or audit worker report
 
 Use this for \`flow-evidence-worker\`, \`flow-reviewer\`,
@@ -47,7 +56,17 @@ success | partial | blocked
 Validation workers must include exact command names and raw outcome summaries
 for commands they actually ran. Audit workers must include guards checked for
 any blocking-severity candidate. Review workers must separate blocking findings
-from advisory notes.
+from advisory notes. In the shared \`Findings or facts\` section, review workers
+should prefix review items with \`blocking:\` or \`advisory:\` before the claim.
+
+Example evidence quality:
+
+- Good fact: \`[high] public Flow command prompts include bundled instructions;
+  evidence: src/config-shared.ts:135; corroboration: single source\`.
+- Weak fact: \`[high] prompts look self-contained; evidence: read the config\`.
+- Good validation: \`bun test tests/distribution-and-surface.test.ts\`, status
+  passed, summary \`surface tests passed and covered bundled command prompts\`.
+- Weak validation: \`tests pass\`, with no command, status, or raw outcome.
 
 ## Verifier worker report
 
@@ -131,6 +150,26 @@ Read these companion references before a broad wave:
 - \`verification-gates.md\` for coverage checks, handoff acceptance, verifier
   triggers, and synthesis rules.
 
+## Operational defaults
+
+- Prefer serial work when the scope is small, tightly coupled, or blocked by one
+  decision that must be made before slices are meaningful.
+- A normal first wave is two to five workers with independent slices. Use more
+  only when the coverage gate is countable and the slices remain non-overlapping.
+- Run at most one routine follow-up wave. Extra waves need an explicit manager
+  reason, such as a high-stakes verifier check or a newly discovered bounded
+  slice.
+- Do not fan out just to keep agents busy. Every worker should reduce a known
+  planning, validation, review, audit, or implementation uncertainty.
+
+Skip fan-out when:
+
+- one file, command, or design question determines the next step.
+- slices would share the same contracts, fixtures, or edit targets.
+- the manager can inspect the full scope faster than writing and checking
+  worker prompts.
+- the result would still need the same manual synthesis with no time saved.
+
 ## Manager sequence
 
 1. Call \`flow_status\` if a Flow session may already exist.
@@ -154,9 +193,10 @@ Read these companion references before a broad wave:
    claims to \`flow-verifier-worker\`.
 9. Run second waves only for material gaps, conflicts, narrowed scope, or
    verification needs.
-10. Synthesize one Flow artifact: plan fields, completion evidence, review
-    payload, audit report, or candidate patch decision. Do not paste worker
-    handoffs as the user-facing result.
+10. Apply the manager synthesis barrier: keep only distilled, evidence-backed
+    claims and synthesize one Flow artifact, such as plan fields, completion
+    evidence, review payload, audit report, or candidate patch decision. Do not
+    paste worker handoffs as the user-facing result.
 
 ## Modes
 
@@ -174,6 +214,36 @@ carry the permission boundaries for each mode.
 | \`verifier\` | \`flow-verifier-worker\` | Per-claim verdicts against cited evidence or commands | No | \`flow_status\` only if needed |
 | \`candidate-implementation\` | \`flow-candidate-worker\` | Candidate patch summary from an isolated worktree or exact path-owned slice | Only with explicit user authorization plus isolation or exact non-overlapping path ownership | No state-changing Flow tools |
 
+Mode examples:
+
+- Use \`flow-evidence-worker\` when the repo shape is unclear and the output will
+  become plan requirements, decisions, targets, or validation entries.
+- Use \`flow-reviewer\` when changed files or risk lenses can be reviewed
+  independently before the manager returns one review payload.
+- Use \`flow-validation-worker\` when the manager needs command options or raw
+  output from an explicitly authorized command.
+- Use \`flow-audit-worker\` when candidate findings must be refuted before they
+  can become a report or follow-up feature.
+- Use \`flow-verifier-worker\` for atomic claims that are contested,
+  single-sourced, high-stakes, or destined for a Flow payload.
+- Use \`flow-candidate-worker\` only after explicit user authorization and only
+  with an isolated worktree or exact non-overlapping path ownership.
+
+## Permission contract
+
+The plugin injects these hidden workers with the following permission values.
+\`Flow state tools\` means the \`flow_*\` rule, while \`Flow status\` documents the
+explicit \`flow_status\` exception.
+
+| Worker | Edit | Bash | Task | Skill | Flow state tools | Flow status |
+| --- | --- | --- | --- | --- | --- | --- |
+| \`flow-reviewer\` | deny | deny | deny | deny | deny | allow |
+| \`flow-evidence-worker\` | deny | deny | deny | deny | deny | allow |
+| \`flow-validation-worker\` | deny | ask | deny | deny | deny | allow |
+| \`flow-audit-worker\` | deny | ask | deny | deny | deny | allow |
+| \`flow-candidate-worker\` | ask | ask | deny | deny | deny | allow |
+| \`flow-verifier-worker\` | deny | ask | deny | deny | deny | allow |
+
 Do not fan out parallel \`flow_plan_save\`, \`flow_plan_approve\`,
 \`flow_run_start\`, \`flow_feature_complete\`, \`flow_feature_reset\`, or
 \`flow_session_close\` calls. Runtime locking protects files, but Flow accepts only
@@ -209,7 +279,7 @@ Mode: evidence | review | validation | audit | verifier | candidate-implementati
 Your exact slice: <paths, modules, command, claim ids, risk lens, or worktree>
 Expected coverage: <count, paths, range, or complete question set>
 Do: <bounded actions>
-Do not: call Flow state tools, edit .flow/**, own sibling slices, or make the final Flow verdict.
+Do not: call state-changing Flow tools, edit .flow/**, own sibling slices, or make the final Flow verdict.
 Return exactly the matching handoff shape from handoff-format.md.
 \`\`\`
 
@@ -217,6 +287,71 @@ For research or current-doc slices, require source checks for versioned or
 time-sensitive facts. For implementation candidates, remind workers that other
 work may be active and that they must not revert unrelated changes.
 
+## Full-wave example
+
+Goal: review whether bundled Flow command guidance is self-contained and aligned
+with hidden worker permissions.
+
+Serial orientation: the manager reads \`src/config-shared.ts\` enough to identify
+five public command templates and six hidden worker configs. The manager keeps
+\`flow-status\` local because it is one line and does not need a worker.
+
+Coverage gate: ten countable items remain after the local check.
+
+- Slice A: \`flow-auto\`, \`flow-plan\`, and \`flow-run\` templates, expected 3/10.
+- Slice B: \`flow-review\` template plus \`flow-reviewer\` config, expected 2/10.
+- Slice C: remaining hidden worker permission blocks, expected 5/10 after
+  excluding the reviewer already covered by Slice B.
+
+Worker prompts:
+
+\`\`\`text
+Overall goal, context only: confirm Flow public commands are self-contained.
+Mode: evidence
+Your exact slice: flow-auto, flow-plan, and flow-run templates in src/config-shared.ts.
+Expected coverage: 3/3 templates.
+Do: report bundled sections, setup preflight coverage, and any gaps with file:line evidence.
+Do not: call state-changing Flow tools, edit .flow/**, own sibling slices, or make the final Flow verdict.
+Return exactly the matching handoff shape from handoff-format.md.
+\`\`\`
+
+\`\`\`text
+Overall goal, context only: confirm Flow review command and hidden reviewer behavior.
+Mode: review
+Your exact slice: flow-review command template and flow-reviewer config in src/config-shared.ts.
+Expected coverage: 2/2 surfaces.
+Do: separate blocking findings from advisory notes and cite file:line evidence.
+Do not: call state-changing Flow tools, edit .flow/**, own sibling slices, or make the final Flow verdict.
+Return exactly the matching handoff shape from handoff-format.md.
+\`\`\`
+
+\`\`\`text
+Overall goal, context only: confirm hidden worker permissions match the orchestration model.
+Mode: audit
+Your exact slice: flow-evidence-worker, flow-validation-worker, flow-audit-worker, flow-candidate-worker, and flow-verifier-worker permissions in src/config-shared.ts.
+Expected coverage: 5/5 worker permission blocks.
+Do: report edit, bash, task, skill, flow_*, and flow_status permissions with evidence.
+Do not: call state-changing Flow tools, edit .flow/**, own sibling slices, or make the final Flow verdict.
+Return exactly the matching handoff shape from handoff-format.md.
+\`\`\`
+
+Handoff checks: the manager accepts only reports with terminal status, matching
+coverage counts, concrete file:line evidence, confidence tags, and claims inside
+the assigned slice. A claim such as \`[high] validation workers may run commands;
+evidence: src/config-shared.ts:281-288; corroboration: single source\` is usable.
+A claim such as \`[high] permissions look safe; evidence: config reviewed\` is
+dropped or retasked.
+
+Verifier pass: the manager sends any single-source claim that will enter the
+Flow payload to \`flow-verifier-worker\`, for example: \`C1: validation, audit,
+candidate, and verifier workers have bash ask while evidence and review workers
+have bash deny; sources: src/config-shared.ts worker permission blocks\`.
+
+Final synthesis: the manager re-reads the relevant config lines, keeps only
+verified or clearly labeled claims, and records one artifact such as a plan
+decision, review payload, or docs patch. Raw handoffs and unverified suggestions
+do not move into the next wave or user-facing answer.
+
 ## Where handoffs go
 
 - Planning evidence becomes \`requirements\`, \`decisions\`, feature \`targets\`,
@@ -235,6 +370,10 @@ work may be active and that they must not revert unrelated changes.
 When worker results conflict, inspect the underlying artifact directly and rerun
 the smallest check that can settle the disagreement.
 
+The manager synthesis barrier means raw handoffs do not move forward by default.
+Only claims that survived coverage, evidence, confidence, and verifier checks may
+enter the next wave, Flow payload, patch decision, or user-facing answer.
+
 ## Second waves
 
 Start a follow-up wave when first-wave handoffs reveal:
@@ -247,7 +386,7 @@ Start a follow-up wave when first-wave handoffs reveal:
 
 Do not recurse by default. If a worker says it needs another worker, the manager
 decides whether that is a second wave and writes the next bounded prompt.
-`;var z='# Recovery playbook\n\nUse this when a Flow tool returns `status: "error"`, a blocker, or a `nextAction` that conflicts with memory.\n\n## First response\n\n1. Re-anchor with `flow_status`.\n2. Read the returned `summary`, `recovery`, `lastError`, and active feature.\n3. Fix the cause, then retry the smallest valid Flow action.\n\n## Common cases\n\n- `missing_session`: start with `flow_plan_save` using the user\'s goal.\n- `missing_goal`: ask for a concrete goal before planning.\n- `Approved plans cannot be changed`: use `flow_feature_reset` when only affected features need another pass; otherwise close and start a new goal.\n- `No feature is currently running`: call `flow_run_start` before completing.\n- `already in progress`: finish, reset, or block the active feature before starting another.\n- `Completion requires recorded validation evidence`: run real validation and include at least one passing `validationRun`.\n- `Completion requires all recorded validation to pass`: fix failures and rerun. Do not relabel failed checks as passed.\n- `Non-final feature completion requires targeted validation`: use `validationScope: "targeted"` for ordinary features.\n- `Final feature completion requires broad validation`: run the project-level gate and use `validationScope: "broad"`.\n- `Completion requires a passing featureReview`: run or request a real review and include a passing `featureReview` only when there are no blocking findings.\n- `Final feature completion requires a finalReview`: perform final review and include `finalReview`.\n- `Final review depth must match the plan policy`: use `reviewDepth` equal to the approved plan\'s `finalReviewPolicy`; valid final-review values are `broad` and `detailed`.\n- `Cannot close ... unfinished features`: complete, reset, defer, or abandon honestly. Do not mark completed while work remains.\n\n## Reset guidance\n\nUse `flow_feature_reset` when the active or completed work was built on the wrong assumption, validation revealed a design issue, dependencies need to be rerun, or dependent features must be invalidated. Resetting a feature also resets its dependents.\n\n## Closure guidance\n\nUse `flow_session_close`:\n\n- `completed`: only after all planned features are complete.\n- `deferred`: the user intentionally postpones unfinished work.\n- `abandoned`: the session should be archived without claiming delivery.\n\nAfter closure, the active `.flow/session.json` is removed and the archived JSON is stored under `.flow/history/`.\n';var B=`# Verification gates
+`;var O='# Recovery playbook\n\nUse this when a Flow tool returns `status: "error"`, a blocker, or a `nextAction` that conflicts with memory.\n\n## First response\n\n1. Re-anchor with `flow_status`.\n2. Read the returned `summary`, `recovery`, `lastError`, and active feature.\n3. Fix the cause, then retry the smallest valid Flow action.\n\n## Common cases\n\n- `missing_session`: start with `flow_plan_save` using the user\'s goal.\n- `missing_goal`: ask for a concrete goal before planning.\n- `Approved plans cannot be changed`: use `flow_feature_reset` when only affected features need another pass; otherwise close and start a new goal.\n- `No feature is currently running`: call `flow_run_start` before completing.\n- `already in progress`: finish, reset, or block the active feature before starting another.\n- `Completion requires recorded validation evidence`: run real validation and include at least one passing `validationRun`.\n- `Completion requires all recorded validation to pass`: fix failures and rerun. Do not relabel failed checks as passed.\n- `Non-final feature completion requires targeted validation`: use `validationScope: "targeted"` for ordinary features.\n- `Final feature completion requires broad validation`: run the project-level gate and use `validationScope: "broad"`.\n- `Completion requires a passing featureReview`: run or request a real review and include a passing `featureReview` only when there are no blocking findings.\n- `Final feature completion requires a finalReview`: perform final review and include `finalReview`.\n- `Final review depth must match the plan policy`: use `reviewDepth` equal to the approved plan\'s `finalReviewPolicy`; valid final-review values are `broad` and `detailed`.\n- `Cannot close ... unfinished features`: complete, reset, defer, or abandon honestly. Do not mark completed while work remains.\n\n## Reset guidance\n\nUse `flow_feature_reset` when the active or completed work was built on the wrong assumption, validation revealed a design issue, dependencies need to be rerun, or dependent features must be invalidated. Resetting a feature also resets its dependents.\n\n## Closure guidance\n\nUse `flow_session_close`:\n\n- `completed`: only after all planned features are complete.\n- `deferred`: the user intentionally postpones unfinished work.\n- `abandoned`: the session should be archived without claiming delivery.\n\nAfter closure, the active `.flow/session.json` is removed and the archived JSON is stored under `.flow/history/`.\n';var B=`# Verification gates
 
 Verification is how Flow keeps parallel work from turning into parallel
 guesswork. Worker handoffs are candidate evidence; the manager decides what can
@@ -322,7 +461,7 @@ Candidate implementation patches are not Flow evidence until the manager
 inspects, merges or rejects them, and runs suitable validation in the main
 Flow-managed workspace.
 
-## Final synthesis
+## Manager synthesis barrier
 
 Before presenting or recording the result:
 
@@ -332,12 +471,14 @@ Before presenting or recording the result:
   instead of arbitrating from summaries.
 - Run the strongest practical local check for the deliverable.
 - Re-read critical files or docs that will be cited in the final decision.
+- Move only distilled, evidence-backed claims forward; raw handoffs remain
+  candidate evidence, not a plan, review, completion payload, or final answer.
 - Record gaps honestly instead of converting missing evidence into success
   language.
 
 \`Status: success\` only says the worker believes its slice is done. The manager
 still checks coverage and evidence before trusting the result.
-`;var O=`---
+`;var z=`---
 name: flow
 description: Run the end-to-end Flow loop for skills-first OpenCode work. Use when a user asks for Flow-guided planning through implementation, resumable autonomous delivery, session status, or completion with validation and review gates.
 ---
@@ -412,7 +553,7 @@ Planning and running require loaded Flow tools; do not simulate plan approval or
 - Unknown runtime error: read \`summary\` and \`recovery\`; see \`references/recovery-playbook.md\` for common cases.
 
 Never fabricate validation output, backfill review approval you did not perform, or close as \`deferred\`/\`abandoned\` merely to avoid an unfinished-work blocker.
-`;var N=`---
+`;var E=`---
 name: flow-commit
 description: Prepare safe Git commits and commit messages. Use only when the user asks to inspect, stage, validate, write a commit message, or create a commit; preserves unrelated work and never pushes, amends, rebases, or publishes without explicit authorization.
 ---
@@ -528,7 +669,7 @@ Before running \`git commit\`, report:
 
 After a successful commit, report the commit hash and leave push or release
 actions for a separate explicit request.
-`;var E=`# Safe refactor workflow
+`;var N=`# Safe refactor workflow
 
 Refactoring is a behavior-preserving sequence of small changes. This workflow keeps cleanup from becoming an unreviewable rewrite.
 
@@ -570,7 +711,7 @@ Weak evidence includes:
 - Public contracts and compatibility shims remain intact or were explicitly planned.
 - Deleted code is actually unreachable or obsolete.
 - Validation can catch a realistic mistake in the refactor.
-`;var V=`# Deslop smell rubric
+`;var G=`# Deslop smell rubric
 
 Use this rubric to turn vague cleanup instincts into reviewable findings.
 
@@ -604,7 +745,7 @@ class; severity; location; evidence read; refutation checked; why it matters; sa
 \`\`\`
 
 Rate as blocking only when the smell materially raises defect risk, blocks planned work, hides behavior, or makes the success claim unverifiable. Style-only cleanup is advisory.
-`;var G=`---
+`;var V=`---
 name: flow-deslop
 description: Clean up and refactor code with evidence-backed code-smell analysis. Use for AI-slop removal, overengineering reduction, maintainability refactors, behavior-preserving cleanup, duplicated or bloated code, speculative abstractions, dead code, or broad cleanup/refactor review.
 ---
@@ -646,7 +787,7 @@ For each claimed smell removal, verify:
 - **blast radius** — public contracts and downstream callers still work.
 
 Never approve cleanup because it "looks cleaner" without evidence. Tests passing is necessary but not sufficient when the refactor changes structure across files.
-`;var L=`# Parallel discovery
+`;var M=`# Parallel discovery
 
 Use this only after a serial orientation pass has identified the repo shape and the likely slices. Workers are read-only evidence gatherers; the planner owns the plan.
 
@@ -679,6 +820,10 @@ For this repository, good first-wave slices are:
   \`bun.lock\`, \`README.md\`, and \`CHANGELOG.md\`.
 - Docs and operator contract: \`docs/**\`, \`README.md\`, and skill references.
 
+Treat these as starting points, not a simultaneous coverage map. Before fan-out,
+choose the relevant entries and de-overlap shared docs, skills, or release
+surfaces in the coverage gate.
+
 ## Coverage gate
 
 Before spawning workers, state the total discovery scope and one line per slice.
@@ -689,17 +834,19 @@ state the completeness rule, such as "all changed files plus callers."
 ## Worker prompt
 
 \`\`\`text
-Inspect <slice> for <goal>. Read-only. Do not edit files or call Flow tools.
-Return the evidence/review/validation/audit handoff shape from ../../flow/references/handoff-format.md.
+Inspect <slice> for <goal>. Read-only. Do not edit files or call
+state-changing Flow tools. Return the evidence/review/validation/audit handoff
+shape from ../../flow/references/handoff-format.md.
 \`\`\`
 
 For validation-oriented discovery:
 
 \`\`\`text
-Inspect <slice> for validation risk. Read-only. Do not edit files or call Flow
-tools. You may report commands that should be run, and include raw output only
-for commands you actually ran. Return the evidence/review/validation/audit
-handoff shape from ../../flow/references/handoff-format.md.
+Inspect <slice> for validation risk. Read-only. Do not edit files or call
+state-changing Flow tools. You may report commands that should be run, and
+include raw output only for commands you actually ran. Return the
+evidence/review/validation/audit handoff shape from
+../../flow/references/handoff-format.md.
 \`\`\`
 
 ## Synthesis
@@ -712,7 +859,11 @@ Convert only evidence-backed work into plan fields:
 - feature \`validation\`: checks expected to prove the feature.
 
 If workers disagree, inspect the source artifact yourself. If a candidate finding lacks a concrete citation or refutation pass, make it a review-first deliverable rather than a fix feature.
-`;var $=`# Planning examples
+
+Apply the manager synthesis barrier from
+\`../../flow/references/verification-gates.md\`: only distilled, evidence-backed
+claims become plan fields.
+`;var L=`# Planning examples
 
 ## Rate limiting feature set
 
@@ -797,7 +948,7 @@ Better plan:
 - Validation that only says "manual testing".
 - Targets that name the entire repo.
 - Features with hidden dependencies instead of \`dependsOn\`.
-`;var M=`---
+`;var $=`---
 name: flow-plan
 description: Plan Flow work for the v4 skills-first runtime: inspect the repo, decompose a user goal into right-sized features, save a draft with flow_plan_save, and approve it with flow_plan_approve.
 ---
@@ -1025,7 +1176,7 @@ Use \`status: "failed"\` when any blocking finding remains. Advisory findings ma
   \`finalReview\` payload.
 
 Never approve to unblock completion, fix findings in the review pass, or vouch for validation you did not inspect.
-`;var X=`# Audit findings rubric
+`;var Y=`# Audit findings rubric
 
 What counts as a valid finding when the feature's deliverable is a findings report: a codebase audit, a review-first feature, or any report whose findings a later feature will fix. The commands you run are still governed by \`validation-rubric.md\`; this rubric governs the findings themselves.
 
@@ -1078,7 +1229,7 @@ follow-up order — correctness and persisted/user-input surfaces first
 \`\`\`
 
 Never: promote a hypothesis to blocking severity; cite a line you did not read in context; rate severity against a deployment model the product does not have; pad the report to look thorough — six verified findings outrank nine where three die on first contact.
-`;var J=`# Validation evidence rubric
+`;var X=`# Validation evidence rubric
 
 Use this before recording \`flow_feature_complete\`.
 
@@ -1135,7 +1286,7 @@ Broad validation usually means the repo's full check command, full relevant test
 - If validation needs external access, missing credentials, or ambiguous user input, record \`status: "needs_input"\` with an honest \`outcome\`.
 
 Never trim failing output, relabel a failed command as passed, or use "not run" as completion evidence.
-`;var Y=`---
+`;var J=`---
 name: flow-run
 description: Execute one approved Flow feature in the v4 runtime: start a feature with flow_run_start, make scoped changes, gather real validation evidence, obtain review payloads, and complete with flow_feature_complete.
 ---
@@ -1478,23 +1629,23 @@ Approve only when the interface is both useful and inspectable:
 - Screenshot/browser evidence supports the claim whenever feasible.
 
 Never approve a UI change based only on code shape. If users will judge it visually, Flow evidence should include visual inspection.
-`;var y=[{name:"flow",files:[{relativePath:"SKILL.md",content:O},{relativePath:"references/recovery-playbook.md",content:z},{relativePath:"references/parallel-orchestration.md",content:W},{relativePath:"references/handoff-format.md",content:S},{relativePath:"references/verification-gates.md",content:B}]},{name:"flow-plan",files:[{relativePath:"SKILL.md",content:M},{relativePath:"references/planning-examples.md",content:$},{relativePath:"references/parallel-discovery.md",content:L}]},{name:"flow-run",files:[{relativePath:"SKILL.md",content:Y},{relativePath:"references/validation-rubric.md",content:J},{relativePath:"references/audit-rubric.md",content:X}]},{name:"flow-test",files:[{relativePath:"SKILL.md",content:Z}]},{name:"flow-review",files:[{relativePath:"SKILL.md",content:K},{relativePath:"references/review-rubric.md",content:Q}]},{name:"flow-deslop",files:[{relativePath:"SKILL.md",content:G},{relativePath:"references/smell-rubric.md",content:V},{relativePath:"references/refactor-workflow.md",content:E}]},{name:"flow-ui-quality",files:[{relativePath:"SKILL.md",content:ae},{relativePath:"references/ui-rubric.md",content:H},{relativePath:"references/visual-verification.md",content:ee}]},{name:"flow-commit",files:[{relativePath:"SKILL.md",content:N}]}];var q=".flow-skill-version";function _(){return process.env.HOME??process.env.USERPROFILE??""}function D(e=_()){return u(e,".config","opencode","skills")}function w(e){return De("sha256").update(e).digest("hex")}function k(e,a){return[`version=${a}`,...e.files.map((t)=>`file=${t.relativePath} sha256=${w(t.content)}`),""].join(`
-`)}async function m(e){try{return await Pe(e,"utf8")}catch(a){if(a.code==="ENOENT")return null;throw a}}function T(e){let a=new Map;if(!e)return a;for(let t of e.split(/\r?\n/)){let r=/^file=(.+) sha256=([a-f0-9]{64})$/.exec(t)??/^file=(.+)=sha256:([a-f0-9]{64})$/.exec(t);if(r?.[1]&&r[2])a.set(r[1],r[2]);let n=/^hash=sha256:([a-f0-9]{64})$/.exec(t);if(n?.[1]&&!a.has("SKILL.md"))a.set("SKILL.md",n[1])}return a}function Oe(e){if(!e)return null;for(let a of e.split(/\r?\n/)){let t=/^version=(.+)$/.exec(a);if(t?.[1])return t[1]}return null}function x(e,a){let t=ze(u(e,...a.split("/")));if(t!==e&&t.startsWith(`${e}${Be}`))return t;throw Error(`Unsafe skill file path '${a}'.`)}async function Ne(e,a){let t=`${e}.backup.${w(a).slice(0,12)}`;for(let r=0;;r+=1){let n=r===0?t:`${t}.${r}`;try{return await b(n,a,{encoding:"utf8",flag:"wx"}),n}catch(s){if(s.code==="EEXIST")continue;throw s}}}async function Ee(e,a,t){let r=u(t,e.name),n=u(r,q),s=await m(n),d=T(s);if(await m(u(r,"SKILL.md"))!==null&&s===null)return{name:e.name,action:"skipped_foreign"};let i=!1,l=[];for(let o of e.files){let c=x(r,o.relativePath),p=await m(c);if(p===o.content)continue;i=!0;let v=d.get(o.relativePath);if(p!==null&&(v?w(p)!==v:s!==null))l.push(await Ne(c,p))}if(!i&&s===k(e,a))return{name:e.name,action:"unchanged"};if(!i)return await b(n,k(e,a),"utf8"),{name:e.name,action:"marker_updated"};let g=s!==null;for(let o of e.files){let c=x(r,o.relativePath);await Te(We(c),{recursive:!0}),await b(c,o.content,"utf8")}return await b(n,k(e,a),"utf8"),{name:e.name,action:l.length>0?"updated_with_backup":g?"updated":"installed",...l.length>0?{backupPaths:l}:{}}}function Ve(){return y.map((e)=>e.name)}function Ge(e){return`npx -y opencode-plugin-flow@${e} doctor`}function I(){if(process.env.npm_package_version)return process.env.npm_package_version;try{let e=Se(import.meta.url);for(let a of["../package.json","../../package.json"])try{let t=e(a);if(t.version)return t.version}catch{}}catch{}return"0.0.0"}async function re(e,a=_()){let t=D(a);return Promise.all(y.map((r)=>Ee(r,e,t)))}async function se(e=I(),a=_()){let t=D(a),r=new Set(Ve()),n=await Promise.all(y.map(async(o)=>{let c=u(t,o.name),p=await m(u(c,q)),v=Oe(p),P=T(p);if(await m(u(c,"SKILL.md"))===null)return{name:o.name,path:c,status:"missing",markerVersion:v,missingFiles:o.files.map((f)=>f.relativePath),editedFiles:[],outdatedFiles:[]};if(p===null)return{name:o.name,path:c,status:"foreign",markerVersion:v,missingFiles:[],editedFiles:[],outdatedFiles:[]};let R=[],C=[],U=[];for(let f of o.files){let A=await m(x(c,f.relativePath));if(A===null){R.push(f.relativePath);continue}if(A===f.content)continue;let j=P.get(f.relativePath);if(j&&w(A)!==j){C.push(f.relativePath);continue}U.push(f.relativePath)}let ce=p!==k(o,e),de=R.length>0?"incomplete":C.length>0?"edited":ce||U.length>0?"outdated":"ok";return{name:o.name,path:c,status:de,markerVersion:v,missingFiles:R,editedFiles:C,outdatedFiles:U}})),s=[];try{s=await oe(t)}catch(o){if(o.code!=="ENOENT")throw o}let d=s.filter((o)=>(o==="flow"||o.startsWith("flow-"))&&!r.has(o)).map((o)=>u(t,o)),h=n.filter((o)=>["missing","incomplete","outdated"].includes(o.status)).map((o)=>o.name),i=n.filter((o)=>["foreign","edited"].includes(o.status)).map((o)=>o.name),l=i.length>0,g=h.length>0;return{status:l?"action_required":g?"sync_required":"ok",version:e,root:t,expectedSkills:[...r],skills:n,syncRequiredSkills:h,actionRequiredSkills:i,unmanagedFlowSkills:d}}function te(e,a,t){if(t.length===0)return;e.push(`- ${a}: ${t.join(", ")}`)}function ne(e){let a=["Flow doctor",`- status: ${e.status}`,`- plugin version: ${e.version}`,`- skills root: ${e.root}`,`- expected skills: ${e.expectedSkills.join(", ")}`];te(a,"startup sync can install/update",e.syncRequiredSkills),te(a,"needs user decision",e.actionRequiredSkills),a.push("","Skills:");for(let t of e.skills){if(a.push(`- ${t.name}: ${t.status} (${t.path})${t.markerVersion?` marker=${t.markerVersion}`:""}`),t.missingFiles.length>0)a.push(`  missing: ${t.missingFiles.join(", ")}`);if(t.editedFiles.length>0)a.push(`  edited: ${t.editedFiles.join(", ")}`);if(t.outdatedFiles.length>0)a.push(`  outdated: ${t.outdatedFiles.join(", ")}`)}if(e.unmanagedFlowSkills.length>0){a.push("","Unmanaged Flow-like skill folders:");for(let t of e.unmanagedFlowSkills)a.push(`- ${t}`)}if(a.push("","Recommendation:"),e.status==="ok")a.push("- Flow skills are present and current.");else if(e.status==="sync_required")a.push("- Start or restart OpenCode with opencode-plugin-flow enabled so startup sync can install or update the listed skills. If Flow then reports restart_required, restart OpenCode once more so the refreshed skill registry is used.");else a.push("- Resolve user-owned or edited managed skill folders, then restart OpenCode. Move a folder aside to let Flow recreate it, or keep it intentionally as a local override.");return a.push(`- Details command: ${Ge(e.version)}`),`${a.join(`
+`;var y=[{name:"flow",files:[{relativePath:"SKILL.md",content:z},{relativePath:"references/recovery-playbook.md",content:O},{relativePath:"references/parallel-orchestration.md",content:W},{relativePath:"references/handoff-format.md",content:S},{relativePath:"references/verification-gates.md",content:B}]},{name:"flow-plan",files:[{relativePath:"SKILL.md",content:$},{relativePath:"references/planning-examples.md",content:L},{relativePath:"references/parallel-discovery.md",content:M}]},{name:"flow-run",files:[{relativePath:"SKILL.md",content:J},{relativePath:"references/validation-rubric.md",content:X},{relativePath:"references/audit-rubric.md",content:Y}]},{name:"flow-test",files:[{relativePath:"SKILL.md",content:Z}]},{name:"flow-review",files:[{relativePath:"SKILL.md",content:K},{relativePath:"references/review-rubric.md",content:Q}]},{name:"flow-deslop",files:[{relativePath:"SKILL.md",content:V},{relativePath:"references/smell-rubric.md",content:G},{relativePath:"references/refactor-workflow.md",content:N}]},{name:"flow-ui-quality",files:[{relativePath:"SKILL.md",content:ae},{relativePath:"references/ui-rubric.md",content:H},{relativePath:"references/visual-verification.md",content:ee}]},{name:"flow-commit",files:[{relativePath:"SKILL.md",content:E}]}];var q=".flow-skill-version";function _(){return process.env.HOME??process.env.USERPROFILE??""}function D(e=_()){return u(e,".config","opencode","skills")}function g(e){return De("sha256").update(e).digest("hex")}function k(e,a){return[`version=${a}`,...e.files.map((t)=>`file=${t.relativePath} sha256=${g(t.content)}`),""].join(`
+`)}async function m(e){try{return await Pe(e,"utf8")}catch(a){if(a.code==="ENOENT")return null;throw a}}function T(e){let a=new Map;if(!e)return a;for(let t of e.split(/\r?\n/)){let r=/^file=(.+) sha256=([a-f0-9]{64})$/.exec(t)??/^file=(.+)=sha256:([a-f0-9]{64})$/.exec(t);if(r?.[1]&&r[2])a.set(r[1],r[2]);let n=/^hash=sha256:([a-f0-9]{64})$/.exec(t);if(n?.[1]&&!a.has("SKILL.md"))a.set("SKILL.md",n[1])}return a}function ze(e){if(!e)return null;for(let a of e.split(/\r?\n/)){let t=/^version=(.+)$/.exec(a);if(t?.[1])return t[1]}return null}function x(e,a){let t=Oe(u(e,...a.split("/")));if(t!==e&&t.startsWith(`${e}${Be}`))return t;throw Error(`Unsafe skill file path '${a}'.`)}async function Ee(e,a){let t=`${e}.backup.${g(a).slice(0,12)}`;for(let r=0;;r+=1){let n=r===0?t:`${t}.${r}`;try{return await b(n,a,{encoding:"utf8",flag:"wx"}),n}catch(s){if(s.code==="EEXIST")continue;throw s}}}async function Ne(e,a,t){let r=u(t,e.name),n=u(r,q),s=await m(n),d=T(s);if(await m(u(r,"SKILL.md"))!==null&&s===null)return{name:e.name,action:"skipped_foreign"};let i=!1,l=[];for(let o of e.files){let c=x(r,o.relativePath),h=await m(c);if(h===o.content)continue;i=!0;let v=d.get(o.relativePath);if(h!==null&&(v?g(h)!==v:s!==null))l.push(await Ee(c,h))}if(!i&&s===k(e,a))return{name:e.name,action:"unchanged"};if(!i)return await b(n,k(e,a),"utf8"),{name:e.name,action:"marker_updated"};let w=s!==null;for(let o of e.files){let c=x(r,o.relativePath);await Te(We(c),{recursive:!0}),await b(c,o.content,"utf8")}return await b(n,k(e,a),"utf8"),{name:e.name,action:l.length>0?"updated_with_backup":w?"updated":"installed",...l.length>0?{backupPaths:l}:{}}}function Ge(){return y.map((e)=>e.name)}function Ve(e){return`npx -y opencode-plugin-flow@${e} doctor`}function I(){if(process.env.npm_package_version)return process.env.npm_package_version;try{let e=Se(import.meta.url);for(let a of["../package.json","../../package.json"])try{let t=e(a);if(t.version)return t.version}catch{}}catch{}return"0.0.0"}async function re(e,a=_()){let t=D(a);return Promise.all(y.map((r)=>Ne(r,e,t)))}async function se(e=I(),a=_()){let t=D(a),r=new Set(Ge()),n=await Promise.all(y.map(async(o)=>{let c=u(t,o.name),h=await m(u(c,q)),v=ze(h),P=T(h);if(await m(u(c,"SKILL.md"))===null)return{name:o.name,path:c,status:"missing",markerVersion:v,missingFiles:o.files.map((f)=>f.relativePath),editedFiles:[],outdatedFiles:[]};if(h===null)return{name:o.name,path:c,status:"foreign",markerVersion:v,missingFiles:[],editedFiles:[],outdatedFiles:[]};let R=[],U=[],C=[];for(let f of o.files){let A=await m(x(c,f.relativePath));if(A===null){R.push(f.relativePath);continue}if(A===f.content)continue;let j=P.get(f.relativePath);if(j&&g(A)!==j){U.push(f.relativePath);continue}C.push(f.relativePath)}let ce=h!==k(o,e),de=R.length>0?"incomplete":U.length>0?"edited":ce||C.length>0?"outdated":"ok";return{name:o.name,path:c,status:de,markerVersion:v,missingFiles:R,editedFiles:U,outdatedFiles:C}})),s=[];try{s=await oe(t)}catch(o){if(o.code!=="ENOENT")throw o}let d=s.filter((o)=>(o==="flow"||o.startsWith("flow-"))&&!r.has(o)).map((o)=>u(t,o)),p=n.filter((o)=>["missing","incomplete","outdated"].includes(o.status)).map((o)=>o.name),i=n.filter((o)=>["foreign","edited"].includes(o.status)).map((o)=>o.name),l=i.length>0,w=p.length>0;return{status:l?"action_required":w?"sync_required":"ok",version:e,root:t,expectedSkills:[...r],skills:n,syncRequiredSkills:p,actionRequiredSkills:i,unmanagedFlowSkills:d}}function te(e,a,t){if(t.length===0)return;e.push(`- ${a}: ${t.join(", ")}`)}function ne(e){let a=["Flow doctor",`- status: ${e.status}`,`- plugin version: ${e.version}`,`- skills root: ${e.root}`,`- expected skills: ${e.expectedSkills.join(", ")}`];te(a,"startup sync can install/update",e.syncRequiredSkills),te(a,"needs user decision",e.actionRequiredSkills),a.push("","Skills:");for(let t of e.skills){if(a.push(`- ${t.name}: ${t.status} (${t.path})${t.markerVersion?` marker=${t.markerVersion}`:""}`),t.missingFiles.length>0)a.push(`  missing: ${t.missingFiles.join(", ")}`);if(t.editedFiles.length>0)a.push(`  edited: ${t.editedFiles.join(", ")}`);if(t.outdatedFiles.length>0)a.push(`  outdated: ${t.outdatedFiles.join(", ")}`)}if(e.unmanagedFlowSkills.length>0){a.push("","Unmanaged Flow-like skill folders:");for(let t of e.unmanagedFlowSkills)a.push(`- ${t}`)}if(a.push("","Recommendation:"),e.status==="ok")a.push("- Flow skills are present and current.");else if(e.status==="sync_required")a.push("- Start or restart OpenCode with opencode-plugin-flow enabled so startup sync can install or update the listed skills. If Flow then reports restart_required, restart OpenCode once more so the refreshed skill registry is used.");else a.push("- Resolve user-owned or edited managed skill folders, then restart OpenCode. Move a folder aside to let Flow recreate it, or keep it intentionally as a local override.");return a.push(`- Details command: ${Ve(e.version)}`),`${a.join(`
 `)}
-`}async function ie(e=_()){let a=D(e),t=[],r=[],n;try{n=await oe(a)}catch(s){if(s.code==="ENOENT")return{removed:t,kept:r};throw s}for(let s of n){if(s!=="flow"&&!s.startsWith("flow-"))continue;let d=u(a,s),h=await m(u(d,q));if(h===null){r.push(d);continue}let i=T(h),l=!1;for(let[g,o]of i){let c=await m(x(d,g));if(c!==null&&w(c)!==o){l=!0;break}}if(l){r.push(d);continue}await je(d,{recursive:!0,force:!0}),t.push(d)}return{removed:t,kept:r}}function F(){return["usage: opencode-plugin-flow <doctor|sync|uninstall> [options]","","commands:","  doctor              Inspect managed Flow skills","  sync                Install or refresh managed Flow skills","  uninstall           Remove pristine Flow-owned managed skills","","doctor options:","  --json              Write the doctor report as JSON","  --check, --strict   Exit nonzero when doctor status is not ok","","global options:","  --help              Show this help","  --version           Print the plugin version"].join(`
-`)}function Le(e,a){return e.every((t)=>a.has(t))}function $e(e,a){if(a.json){process.stdout.write(`${JSON.stringify(e,null,2)}
-`);return}process.stdout.write(ne(e))}async function Me(e){let a=e[2],t=e.slice(3);if(a==="--help"||a==="-h"){process.stdout.write(`${F()}
+`}async function ie(e=_()){let a=D(e),t=[],r=[],n;try{n=await oe(a)}catch(s){if(s.code==="ENOENT")return{removed:t,kept:r};throw s}for(let s of n){if(s!=="flow"&&!s.startsWith("flow-"))continue;let d=u(a,s),p=await m(u(d,q));if(p===null){r.push(d);continue}let i=T(p),l=!1;for(let[w,o]of i){let c=await m(x(d,w));if(c!==null&&g(c)!==o){l=!0;break}}if(l){r.push(d);continue}await je(d,{recursive:!0,force:!0}),t.push(d)}return{removed:t,kept:r}}function F(){return["usage: opencode-plugin-flow <doctor|sync|uninstall> [options]","","commands:","  doctor              Inspect managed Flow skills","  sync                Install or refresh managed Flow skills","  uninstall           Remove pristine Flow-owned managed skills","","doctor options:","  --json              Write the doctor report as JSON","  --check, --strict   Exit nonzero when doctor status is not ok","","global options:","  --help              Show this help","  --version           Print the plugin version"].join(`
+`)}function Me(e,a){return e.every((t)=>a.has(t))}function Le(e,a){if(a.json){process.stdout.write(`${JSON.stringify(e,null,2)}
+`);return}process.stdout.write(ne(e))}async function $e(e){let a=e[2],t=e.slice(3);if(a==="--help"||a==="-h"){process.stdout.write(`${F()}
 `);return}if(a==="--version"||a==="-v"){process.stdout.write(`${I()}
 `);return}if(a!=="uninstall"&&a!=="doctor"&&a!=="sync"){process.stderr.write(`${F()}
-`),process.exitCode=2;return}if(a==="doctor"){if(!Le(t,new Set(["--json","--check","--strict"]))){process.stderr.write(`${F()}
-`),process.exitCode=2;return}let s=await se();if($e(s,{json:t.includes("--json")}),(s.status==="sync_required"||s.status==="action_required")&&(t.includes("--check")||t.includes("--strict")))process.exitCode=1;return}if(t.length>0){process.stderr.write(`${F()}
-`),process.exitCode=2;return}if(a==="sync"){let n=I(),s=await re(n),d=s.filter((i)=>["installed","updated","updated_with_backup"].includes(i.action)),h=s.filter((i)=>i.action==="skipped_foreign");process.stdout.write(`Flow skill sync (${n})
+`),process.exitCode=2;return}if(a==="doctor"){if(!Me(t,new Set(["--json","--check","--strict"]))){process.stderr.write(`${F()}
+`),process.exitCode=2;return}let s=await se();if(Le(s,{json:t.includes("--json")}),(s.status==="sync_required"||s.status==="action_required")&&(t.includes("--check")||t.includes("--strict")))process.exitCode=1;return}if(t.length>0){process.stderr.write(`${F()}
+`),process.exitCode=2;return}if(a==="sync"){let n=I(),s=await re(n),d=s.filter((i)=>["installed","updated","updated_with_backup"].includes(i.action)),p=s.filter((i)=>i.action==="skipped_foreign");process.stdout.write(`Flow skill sync (${n})
 `);for(let i of s){process.stdout.write(`- ${i.name}: ${i.action}
 `);for(let l of i.backupPaths??[])process.stdout.write(`  backup: ${l}
 `)}if(d.length>0)process.stdout.write(`Restart OpenCode so the refreshed skill registry is used.
-`);if(h.length>0)process.stdout.write(`Some managed skill folders are user-owned or edited; run doctor for repair guidance.
+`);if(p.length>0)process.stdout.write(`Some managed skill folders are user-owned or edited; run doctor for repair guidance.
 `);return}let r=await ie();for(let n of r.removed)process.stdout.write(`Removed Flow skill: ${n}
 `);for(let n of r.kept)process.stdout.write(`Kept non-Flow or user-edited skill: ${n}
 `);process.stdout.write(`Remove opencode-plugin-flow from your OpenCode plugin config and restart OpenCode.
-`)}Me(process.argv).catch((e)=>{process.stderr.write(`${e instanceof Error?e.message:String(e)}
+`)}$e(process.argv).catch((e)=>{process.stderr.write(`${e instanceof Error?e.message:String(e)}
 `),process.exitCode=1});