Add /improve-skills: meta-critic that improves the system's own prompts

.claude/commands/dream.md

@@ -36,7 +36,10 @@ $ARGUMENTS
    verifiable bullet. If it fits no existing page, create a new
    `topics/<slug>.md`, add cross-links, and add one row to `index.md`.
 5. If a learning is really a session-governing rule (a do/don't policy or a
-   completion gate), propose it for `CLAUDE.md` and apply the edit.
+   completion gate), it belongs in `CLAUDE.md`. Show the proposed `CLAUDE.md` edit and
+   apply it **only after explicit confirmation** — `CLAUDE.md` is a core governing
+   prompt (it overrides every session), so it always needs confirmation, even for an
+   addition. Topic-page and index-row additions do not.
 6. **Skills check.** Decide whether the learning is a *fact* or a *recurring
    procedure*:
    - A **fact** (gotcha, convention, command) → a topic page, as above.
@@ -83,6 +86,8 @@ $ARGUMENTS
 - Never record temporary or branch-specific bugs, or one-time errors already fixed.
 - Prefer appending; editing or removing an existing fact needs explicit user
   confirmation. Adding a topic or an index row does not.
+- Adding or changing a rule in `CLAUDE.md` (a core governing prompt), unlike adding a
+  topic-page fact, **always** needs explicit confirmation.
 - Never auto-create, rewrite, or delete a command/skill. Propose it and act only on
   confirmation. Updating `.claude/skills-index.md` rows additively is fine.
 - Keep `index.md` to summaries + links; facts belong in topic pages.

.claude/commands/fix-localhost.md

@@ -11,21 +11,25 @@ Success criteria:
 - No blocking browser console errors on initial load.
 - No failed required network requests on initial load.
 - Typecheck, lint, and relevant tests pass.
-- Post-fix critic round has no blockers.
+- Post-fix single-agent self-review has no blockers.
 
 Procedure:
 
 1. Run the dev server.
 2. Open localhost with Chrome or Playwright MCP.
 3. Capture the exact browser/terminal error.
-4. Run this critic round:
-   - First-Time User Critic
-   - Frontend Architecture Critic
-   - QA / E2E Critic
-   - Regression Critic
+4. As Lead Agent, self-review the failure inline through these four lenses — this
+   command is intentionally **single-agent**: do NOT fan out to the critic-panel
+   Workflow. (If a full critic round is warranted, invoke `/critic-round` separately.)
+   List each finding with severity + evidence:
+   - First-Time User
+   - Frontend Architecture
+   - QA / E2E
+   - Regression
 5. Diagnose root cause.
 6. Apply the smallest safe fix.
 7. Re-test in browser.
 8. Run typecheck, lint, and tests.
-9. Run a post-fix critic round.
+9. Repeat the single-agent self-review (step 4) after the fix; if the fix is
+   non-trivial, run `/critic-round` (the critic-panel Workflow) instead.
 10. Summarize evidence.

.claude/commands/improve-skills.md

@@ -0,0 +1,54 @@
+---
+description: Meta-critic pass over the project's OWN prompts, skills, and workflow orchestration — fan out one reviewer per quality lens, adversarially verify each finding, and PROPOSE edits to the core prompts. Never auto-applies; every change is confirmation-gated. Use periodically to let the system improve its own machinery, not just its facts.
+argument-hint: [scope, e.g. "the loop workflows" | "all commands"]
+disable-model-invocation: true
+---
+
+Run a **self-improvement pass over the system itself** — the custom prompts, skills,
+and multi-agent orchestration that drive this project — not over product code. This
+is the counterpart to `/dream`: `/dream` improves what the system *knows* (the wiki);
+`/improve-skills` improves the *machinery that does the work* (the command and
+workflow prompts), by dogfooding the same critic-panel pattern on our own config.
+
+Scope:
+$ARGUMENTS
+
+## Procedure
+1. **Run the meta-critic as a deterministic Workflow** — invoke the Workflow tool with
+   `scriptPath: .claude/workflows/improve-skills.js`,
+   `args: { "scope": "$ARGUMENTS" }` (optionally pass `args.targets` = a specific list
+   of files to scope the review). One reviewer per quality lens (clarity,
+   instruction-following risk, safety/gating, consistency/DRY, workflow DSL, routing)
+   runs in its own context, in parallel; a *separate* skeptic adversarially verifies
+   every blocker/important finding. Invoking Workflow here is expected.
+2. The workflow returns `{ scope, targets, counts, applyReady, needsDesign, niceToHaves, refuted }`.
+   Render `applyReady` and `needsDesign` as a matrix and list `niceToHaves` separately:
+
+   | Lens | Severity | File | Finding | Evidence | Proposed edit | Edit verified safe? |
+   |---|---|---|---|---|---|---|
+
+3. **Confirmation gate (hard rule).** This command may propose edits to core prompts
+   but must **never apply them without explicit confirmation** — the same rule
+   `/dream` follows for creating/rewriting skills. Present the proposed edits and ask
+   before touching any file.
+4. After confirmation, apply only the approved `applyReady` edits (the ones whose
+   `editSafe` is true). Treat `needsDesign` findings (real problem, edit not verified
+   safe) as items needing a human-designed fix — surface them, don't auto-apply.
+5. Do **not** weaken an instruction, remove a safety/confirmation gate, or delete a
+   command/skill as part of an "improvement". Removing or rewriting an existing rule
+   needs explicit confirmation; additive clarifications are lower-risk but still shown
+   before applying.
+6. When an edit changes a command or workflow, keep `.claude/skills-index.md` and any
+   affected memory wiki page in sync (additive index updates need no confirmation).
+7. End the task with `/dream` so any durable lesson from this pass is filed into the
+   wiki.
+
+## Notes
+- Heavier, deterministic path: the workflow spawns ~6 meta-critics + one verifier per
+  blocker/important finding. That is intended — independent review of our own prompts,
+  not one context grading its own instructions.
+- The verifier judges two things: is the problem **real**, and is the proposed edit
+  **safe** (would it weaken instruction-following or a safety gate, or break a
+  workflow?). Only real-and-safe findings are `applyReady`.
+- `disable-model-invocation: true`: this command changes core prompts, so it is
+  manual-only — never auto-invoked from a description match.

.claude/commands/multi-agent-dev.md

@@ -44,6 +44,7 @@ Never mark complete unless:
 
 - the flow works in browser
 - no blocking console errors remain
+- typecheck, lint, and unit tests pass (build passes when the change warrants it)
 - relevant E2E tests pass
 - post-implementation critic round has no blockers
 - files changed are reported

.claude/commands/write-goal.md

@@ -15,7 +15,8 @@ $ARGUMENTS
    `.claude/memory/index.md` → the relevant topic page(s), so the prompt uses this
    project's **real** commands, conventions, and gates — not generic ones.
 2. If the task is broad or unclear, ask 1–2 sharp clarifying questions first, or
-   suggest running `/plan` and converting the plan into the goal.
+   first outline the plan inline (the steps + files it touches) and convert that
+   outline into the goal.
 3. Produce a single, copy-pasteable prompt in a fenced block, grounded in this repo.
 
 ## A `/goal` must include all six

.claude/loop.md

@@ -16,8 +16,11 @@ budget, denylist, escalation, stop conditions). Before running, fix the loop's
 1. **Read state first:** open `.claude/memory/loop-run-log.md` (prior iterations,
    outcomes, human overrides) and check **remaining budget** against the per-run cap
    (turns or tokens — see *Budget & limits*) before doing anything; stop and
-   escalate if it is already exhausted.
-2. Identify the current active task from the conversation.
+   escalate if it is already exhausted. **If no per-run cap was declared, default to
+   25 turns and record that you used the default** in the run-log entry.
+2. Identify the current active task from the conversation or run-log. **If no active
+   task is identifiable, treat the watchlist as empty and self-stop (per *Stop when*)
+   rather than inventing a task.**
 3. Check git status and changed files.
 4. Start or reuse the dev server.
 5. Open the relevant localhost page in Chrome or Playwright MCP.
@@ -28,7 +31,8 @@ budget, denylist, escalation, stop conditions). Before running, fix the loop's
    Render the returned `confirmed` matrix.
 8. **Maker/checker:** the pass that verifies a fix must be *separate* from the one
    that made it — the implementer never marks its own work "done".
-9. Act **only within the trust level**:
+9. Act **only within the trust level** (**if no level was explicitly declared for
+   this run, treat it as L1 — never infer L2/L3 from context**):
    - L1: record findings only.
    - L2/L3: fix the confirmed blockers + directly-related important findings via the
      **loop-iteration Workflow** (`.claude/workflows/loop-iteration.js`,
@@ -43,7 +47,12 @@ budget, denylist, escalation, stop conditions). Before running, fix the loop's
 ## Denylist — never touch autonomously (escalate instead)
 Auth, payments, secrets/`.env`, infrastructure/deploy, CI workflow config, and
 database migrations. Also: do not push, deploy, delete data, modify secrets, or
-auto-merge without an explicit allowlist.
+auto-merge without an explicit allowlist. **And the agent's own configuration and
+prompt surface — `.claude/` (commands, workflows, `loop.md`, `loop-checklist.md`,
+the memory wiki) and `CLAUDE.md`: a loop never self-modifies its own instructions
+autonomously, even at L3 — propose and escalate instead** (this is enforced
+deterministically by the `loop-iteration.js` denylist regex; the sanctioned path to
+change the prompt surface is the manual, confirmation-gated `/improve-skills`).
 
 ## Escalate to a human (pause and ask) when
 - A fix would touch a denylist path.

.claude/memory/log.md

@@ -113,3 +113,41 @@ change), `rotate` (archived old entries). Format: `## [YYYY-MM-DD] <op> | <summa
 - Lesson: a maker/checker that hands the checker only the maker's self-report is
   weak — give the checker the real artifact (diff). Exact-title dedup also let one
   bug appear as two findings; merge, don't drop.
+
+## [2026-06-29] ingest | Self-improvement axis: /improve-skills meta-critic
+- The self-learning system improved only its *facts* (the wiki via `/dream`), not its
+  *own prompts/workflows*. Added `/improve-skills` (`.claude/workflows/improve-skills.js`
+  + `.claude/commands/improve-skills.md`): a per-lens meta-critic (clarity,
+  instruction-following risk, safety/gating, consistency/DRY, workflow DSL, routing)
+  over the project's OWN command/workflow/instruction surface, with a separate skeptic
+  verifying each finding on two axes (problem **real** + edit **safe**). Returns
+  proposed edits only — confirmation-gated, `disable-model-invocation: true`.
+- Wired into CLAUDE.md (workflow list + "self-improvement vs self-learning" note),
+  skills-index, and topics/workflow.md. Filed the two-axis principle there.
+
+## [2026-06-29] ingest | First /improve-skills run found a self-modification gap
+- Dogfooded `/improve-skills` (25 agents, 6 lenses): 2 blockers + 10 important
+  confirmed, 7 refuted (the two-axis verifier rejected 3 "improvements" that would
+  have *weakened* a gate — proof the editSafe axis works). Applied 7 fixes:
+  - **Self-modification guard (blocker):** the autonomous-loop denylist protected only
+    product/infra paths, so an L2/L3 loop could rewrite its OWN governing prompts.
+    Added `.claude/` + `CLAUDE.md` to the `loop-iteration.js` regex gate AND the
+    `loop.md` prose denylist. The sanctioned path to edit prompts is the manual,
+    confirmation-gated `/improve-skills`.
+  - `/dream`: gated `CLAUDE.md` edits behind explicit confirmation (was "propose and
+    apply" — contradicted its own Hard rules); added the matching Hard rule.
+  - `CLAUDE.md`: added "commit/push only when the user asks" (settings pre-allow
+    `git commit`, but no rule said not to commit unprompted).
+  - `loop.md`: fail-safe defaults — unstated per-run cap → 25 turns, unstated trust
+    level → L1 at the action step, no identifiable task → self-stop (don't invent one).
+  - `fix-localhost`: made its single-agent intent explicit (it was naming a critic
+    panel inline, which CLAUDE.md says must be a Workflow / is single-agent).
+  - `multi-agent-dev`: its "never mark complete unless" gate omitted
+    typecheck/lint/unit/build that CLAUDE.md requires — added them.
+  - `write-goal` + `CLAUDE.md`: removed the dangling `/plan` reference (no such repo
+    command) → "outline the plan inline".
+  - Hardened `improve-skills.js` to parse a JSON-string `args` (a command may pass
+    args as a string; was leaking the raw JSON into `scope`).
+- Lesson: a self-improving system MUST add its own prompt surface (`.claude/`,
+  `CLAUDE.md`) to every autonomous denylist — otherwise the loop can edit the rules
+  that constrain it. The meta-critic caught this in its own machinery on the first run.

.claude/memory/topics/workflow.md

@@ -23,11 +23,28 @@
   `.claude/workflows/`, not one agent simulating critics: `critic-panel` (critics →
   adversarial verify → matrix), `scan-docs` (per-story evidence → status verifier),
   `gap-analysis` (per-dimension gaps → synth order), `e2e-design` (per-category
-  cases → dedup), `loop-iteration` (worktree implementer → separate verifier).
-  Commands invoke them via the Workflow tool (the sanctioned opt-in). Name a workflow
+  cases → dedup), `loop-iteration` (worktree implementer → separate verifier),
+  `improve-skills` (per-lens meta-critic over our OWN prompts/workflows → proposed
+  edits, confirmation-gated). Commands invoke them via the Workflow tool (the
+  sanctioned opt-in). Name a workflow
   distinctly from any command (e.g. `critic-panel` vs the `/critic-round` command) to
   avoid a duplicate skills-list entry. Don't force non-fan-out commands
   (`/fix-localhost`, `/write-goal`, `/dream`, `/memory-audit`) into workflows.
+- **Self-improvement vs self-learning (two axes).** `/dream` improves what the system
+  *knows* (curates the memory wiki — facts); `/improve-skills` improves the *machinery*
+  (a meta-critic Workflow that reviews the project's OWN command/workflow prompts and
+  proposes tightenings). Both are confirmation-gated: never auto-create, rewrite, or
+  delete a command/skill, and never weaken an instruction or remove a safety gate as
+  an "improvement". `/improve-skills` is `disable-model-invocation: true` (manual-only)
+  because it edits core prompts; its verifier judges a finding on *two* axes — is the
+  problem **real** and is the proposed edit **safe** (won't weaken instruction-following
+  or break a workflow) — and only real-and-safe findings are `applyReady`.
+- **A self-improving system must add its OWN prompt surface to every autonomous
+  denylist.** The loop's deterministic gate (`loop-iteration.js` DENYLIST) and prose
+  denylist (`loop.md`) both include `.claude/` and `CLAUDE.md`, so an L2/L3 loop can
+  never autonomously rewrite the rules that constrain it — it escalates instead. The
+  only sanctioned way to change the prompt surface is the manual, confirmation-gated
+  `/improve-skills`. (Found by the first `/improve-skills` run, on its own machinery.)
 - Workflow scripts use a DSL (`agent()/parallel()/pipeline()/phase()`, top-level
   `await`/`return`); validate syntax by wrapping the body in `async function(){…}`
   before `node --check` (bare `node --check` reports a false "illegal return").

.claude/skills-index.md

@@ -22,6 +22,7 @@ catalog so both you and Claude can route to the right command.
 - **Map what's built vs documented vs tested** → `/scan-project-docs`, then
   `/sync-story-status`, `/story-gap-analysis`
 - **Capture/curate what was learned** → `/dream`, `/memory-audit`
+- **Improve the system's own prompts/skills/workflows** → `/improve-skills`
 - **Keep watching localhost on an interval** → `/loop` (uses `.claude/loop.md`;
   gate trust level with `.claude/loop-checklist.md`, state in
   `.claude/memory/loop-run-log.md`)
@@ -32,7 +33,8 @@ catalog so both you and Claude can route to the right command.
 > (parallel subagents + adversarial/maker-checker verification), invoked via the
 > Workflow tool — not one agent simulating critics: `critic-panel` (critic rounds),
 > `scan-docs` (`/scan-project-docs`), `gap-analysis` (`/story-gap-analysis`),
-> `e2e-design` (`/multi-agent-e2e`), `loop-iteration` (`/loop` L2/L3).
+> `e2e-design` (`/multi-agent-e2e`), `loop-iteration` (`/loop` L2/L3),
+> `improve-skills` (`/improve-skills` — meta-critic over our own prompts/workflows).
 
 | Command | When to use | Args |
 |---|---|---|
@@ -55,6 +57,7 @@ catalog so both you and Claude can route to the right command.
 |---|---|---|
 | `/dream` | End of a non-trivial task: ingest durable learnings into the wiki, lint it, and propose new skills | `[ingest \| query <q> \| lint]` |
 | `/memory-audit` | Periodically prune stale/duplicated/misleading memory (no deletes without confirmation) | — |
+| `/improve-skills` | Periodically improve the system's OWN prompts/skills/workflows via a meta-critic; proposes edits, confirmation-gated (manual-only) | `[scope]` |
 
 ## Conventions for adding a command
 - Every command needs a one-line `description` frontmatter (used for routing) and,