@event4u/agent-config 16.2.0 → 16.4.0

package/README.md

@@ -4,9 +4,9 @@
 
 [![Smoke](https://github.com/event4u-app/agent-config/actions/workflows/smoke.yml/badge.svg)](https://github.com/event4u-app/agent-config/actions/workflows/smoke.yml) [![Public install smoke (3 OS × 2 Node)](https://github.com/event4u-app/agent-config/actions/workflows/smoke-public-install.yml/badge.svg)](https://github.com/event4u-app/agent-config/actions/workflows/smoke-public-install.yml) [![npm](https://img.shields.io/npm/v/@event4u/agent-config?style=flat-square&label=npm&color=orange)](https://www.npmjs.com/package/@event4u/agent-config) [![agent-config MCP server](https://glama.ai/mcp/servers/event4u-app/agent-config/badges/score.svg)](https://glama.ai/mcp/servers/event4u-app/agent-config) [![MCP Toplist](https://mcptoplist.com/badge/glama%2Fevent4u-app%2Fagent-config.svg)](https://mcptoplist.com/server/glama%2Fevent4u-app%2Fagent-config)
 
-[![Skills](https://img.shields.io/badge/Skills-299-orange?style=flat-square)](dist/agent-src/skills/) [![Rules](https://img.shields.io/badge/Rules-121-orange?style=flat-square)](dist/agent-src/rules/) [![Commands](https://img.shields.io/badge/Commands-203-orange?style=flat-square)](dist/agent-src/commands/) [![Guidelines](https://img.shields.io/badge/Guidelines-121-orange?style=flat-square)](docs/guidelines/) [![Personas](https://img.shields.io/badge/Personas-29-orange?style=flat-square)](dist/agent-src/personas/) [![Advisors](https://img.shields.io/badge/Advisors-5-orange?style=flat-square)](dist/agent-src/personas/advisors/) <!-- claim:published-artifact-counts -->
+[![Skills](https://img.shields.io/badge/Skills-299-orange?style=flat-square)](dist/agent-src/skills/) [![Rules](https://img.shields.io/badge/Rules-122-orange?style=flat-square)](dist/agent-src/rules/) [![Commands](https://img.shields.io/badge/Commands-205-orange?style=flat-square)](dist/agent-src/commands/) [![Guidelines](https://img.shields.io/badge/Guidelines-121-orange?style=flat-square)](docs/guidelines/) [![Personas](https://img.shields.io/badge/Personas-29-orange?style=flat-square)](dist/agent-src/personas/) [![Advisors](https://img.shields.io/badge/Advisors-5-orange?style=flat-square)](dist/agent-src/personas/advisors/) <!-- claim:published-artifact-counts -->
 
-<sub>**How these are counted** — one canonical counter, `agent-config` → `update_counts --check`, re-derives all six from the tree and fails CI on a drift of one. Two bases are not what the linked directory shows, so they are stated here rather than left to be inferred: **Commands 202** counts every command file **recursively** (the linked directory holds 61 at its top level), and **Rules 121** counts the **source** rules while the linked projection holds 120 — one rule is dormant and is not projected. **Personas 29** excludes the directory README. Counts are of files and directories: none of them measures quality, activation, or adoption. <!-- claim:published-artifact-counts --></sub>
+<sub>**How these are counted** — one canonical counter, `agent-config` → `update_counts --check`, re-derives all six from the tree and fails CI on a drift of one. Two bases are not what the linked directory shows, so they are stated here rather than left to be inferred: **Commands 205** counts every command file **recursively** (the linked directory holds 60 at its top level), and **Rules 121** counts the **source** rules while the linked projection holds 120 — one rule is dormant and is not projected. **Personas 29** excludes the directory README. Counts are of files and directories: none of them measures quality, activation, or adoption. <!-- claim:published-artifact-counts --></sub>
 
 **Try it in 30 seconds** — drop one read-only subagent into any repo and watch it gate "done": `@production-validator check this branch is actually done`. No wizard, no lock-in, nothing else installed — the [30-second wedge ↓](#quickstart) is the whole first step. Start at the proof, not the catalog: **[event4u-app.github.io/agent-config/proof/](https://event4u-app.github.io/agent-config/proof/)**.
 
@@ -20,6 +20,18 @@
 
 ## Try it in 30 seconds
 
+```bash
+# 1. Install — on a terminal with a display, the browser wizard launches
+#    automatically; the same TypeScript installer runs the real install behind it.
+npx -y @event4u/agent-config init
+
+# 2. Pick your profile + tools in the wizard, click Finish.
+#    (Writes ~/.event4u/agent-config/, ~/.claude/, ~/.cursor/, …)
+
+# 3. First real task — agent refines, plans, verifies.
+/work "your first real task"
+```
+
 **Try one thing in 30 seconds** — before the full suite, drop in a single
 self-contained subagent and see the discipline on your own repo:
 
@@ -165,17 +177,7 @@ How *long* it takes is not: that is dominated by network and registry latency.
 CI measures the install → `doctor` wall-clock on every umbrella run and
 publishes it with its conditions, as evidence rather than as a promise.
 
-```bash
-# 1. Install — on a terminal with a display, the browser wizard launches
-#    automatically; the same TypeScript installer runs the real install behind it.
-npx -y @event4u/agent-config init
-
-# 2. Pick your profile + tools in the wizard, click Finish.
-#    (Writes ~/.event4u/agent-config/, ~/.claude/, ~/.cursor/, …)
-
-# 3. First real task — agent refines, plans, verifies.
-/work "your first real task"
-```
+**The one command is [above ↑](#try-it-in-30-seconds).**
 
 **Headless / CI:** `init` skips the GUI on CI, a non-TTY, a headless host, or any CLI-mode flag, and runs the non-interactive installer instead. The GUI and the CLI share one installer (`src/scripts/install.ts`), so both produce identical results. Flags, the full opt-out set and `--dry-run`: [`docs/wizard.md`](docs/wizard.md) · [`gui-wizard` § When the GUI is skipped](docs/contracts/gui-wizard.md#when-the-gui-is-skipped).
 

package/agents/templates/.ai-council.yml.example

@@ -29,9 +29,14 @@
 # `api` rung. A vendor CLI under your subscription is preferred whenever one
 # resolves, so the common case costs nothing per token — run
 # `agent-config council status` to see the resolved transport and billing class
-# per member before invoking. The autonomy directive does NOT silently spend
-# tokens — the /council command always asks before invoking, even under
-# autonomy: on.
+# per member before invoking.
+#
+# This block used to end "the /council command always asks before invoking,
+# even under autonomy: on". That was already untrue when ADR-230 replaced the
+# unconditional ask with a three-case bound, and ADR-279 has since removed the
+# last case that asked. The command does not ask; it shows the estimate as
+# information. What bounds a billable run is a ceiling you set below — and
+# with nothing set, nothing bounds it.
 
 # Master switch (true, false). REQUIRED — the loader has no default for it,
 # so this file's value is the only value.
@@ -138,11 +143,21 @@ defaults:
 # Prices come from agents/.agent-prices.md (gitignored, refreshed weekly
 # by `python3 scripts/update_prices.py`; bootstrapped from
 # scripts/ai_council/_default_prices.py on first run).
+#
+# These ship at 0 — no ceiling on tokens or money unless you set one
+# (ADR-279). `max_calls` is NOT a ceiling on money: it bounds fan-out per
+# invocation, so it keeps its value. To get a bound back, replace a 0 with the
+# figure you want; the units are TOKENS for the first two and US DOLLARS for
+# `max_total_usd`.
+#
+# If you copied this file before 2026-10-06 it carries 500000 / 200000 / 20.0
+# explicitly. Those are YOUR figures now and nothing rewrites them — delete the
+# lines to become unbounded, or keep them to stay bounded.
 cost_budget:
-  max_input_tokens: 500000
-  max_output_tokens: 200000
+  max_input_tokens: 0
+  max_output_tokens: 0
   max_calls: 50
-  max_total_usd: 20.0
+  max_total_usd: 0.0
 
 # Phase 8 — Debate cost-visibility + hard refusal cap.
 #
@@ -150,7 +165,10 @@ cost_budget:
 # high-end estimate (expected × 1.20 over-run buffer) exceeds the cap,
 # `/council debate` refuses to start — the user must lower `--rounds`,
 # drop members, or raise the cap explicitly. No --proceed-anyway override.
-# Set to 0 to disable the cap (not recommended).
+#
+# It ships at 0 — disabled, per ADR-279. Set a USD figure to get the refusal
+# back. The between-round confirmation below is a separate control and is NOT
+# a money ceiling: it stays on whatever this is set to.
 #
 # `cost_disclosure.mode` ∈ {"always", "above_threshold", "off"}:
 #   always           — emit the disclosure block before every debate.
@@ -165,7 +183,7 @@ cost_budget:
 #         mode: above_threshold
 #         threshold_usd: 0.50
 debate:
-  max_cost_usd: 5.00
+  max_cost_usd: 0.00
   cost_disclosure:
     mode: always
     threshold_usd: 1.00

package/dist/agent-src/commands/analyze/inbox.md

@@ -672,13 +672,16 @@ attempt. At the third, write the row with what was observed and move on.
 
 This cap is **this phase's own**, and the divergence from
 [`autonomous-execution`](../../../../rules/autonomous-execution.md)
-§ Validation-loop budget is deliberate rather than an oversight: that rule's N=3
-governs a *validation target* it defines as "a single identifiable artefact (file
-path, lint rule ID, test name)", explicitly excluding natural-language clusters —
-and a verbatim step out of a prose file is exactly such a cluster. Its remedy is
-also different (STOP and ask the user). Borrowing the number while failing the
-definition, and then citing the rule as authority, would weaken the rule it
-claims to obey. So: same number, stated here, for a reason stated here.
+§ Validation-loop budget is deliberate rather than an oversight: that rule's
+`execution.fix_loop_max` budget (default 10, with a strategy-shift ladder)
+governs a *validation target* it defines as "a single identifiable artefact
+(file path, lint rule ID, test name)", explicitly excluding natural-language
+clusters — and a verbatim step out of a prose file is exactly such a cluster.
+This phase's own remedy is also different: STOP and ask the user, where the
+rule's remedy at the same count is a mandatory strategy shift, never a stop.
+Borrowing the number while failing the definition, and then citing the rule
+as authority, would weaken the rule it claims to obey. So: same number,
+stated here, for a reason stated here.
 
 One row per selected step:
 
@@ -1189,7 +1192,7 @@ Declared by `produces_roadmap: true` in the frontmatter and enforced by
 `lint_roadmap_producers`: a producer that does not end here reds CI.
 
 Read `planning.closure_pass` (missing = `true`). When active, run
-[`/challenge-me closure`](../../../meta/challenge-me/closure/command.md) on the
+`/challenge-me closure` on the
 roadmap this command just produced, before handing back:
 
 ```bash

package/dist/agent-src/commands/analyze/roadmap-repos.md

@@ -140,7 +140,7 @@ Declared by `produces_roadmap: true` in the frontmatter and enforced by
 `lint_roadmap_producers`: a producer that does not end here reds CI.
 
 Read `planning.closure_pass` (missing = `true`). When active, run
-[`/challenge-me closure`](../../../meta/challenge-me/closure/command.md) on the
+`/challenge-me closure` on the
 roadmap this command just produced, before handing back:
 
 ```bash

package/dist/agent-src/commands/bug/fix.md

@@ -5,14 +5,13 @@ pack: engineering-base
 visibility: internal
 sub: fix
 cluster: bug
-skills: [bug-analyzer, pest-testing]
+skills: [bug-analyzer, test-case-discovery]
 description: Plan and implement a bug fix — based on investigation, with quality checks and test verification
 argument-hint: "[bug-description]"
 council_depth: deep
 suggestion:
   eligible: false
   rationale: "Cluster sub-command — reached via its cluster head's routing or its explicit /cluster:sub name; not independently suggested (surface-consolidation)."
-framework: laravel
 workspaces:
   - agent-config-maintainer
 packs:
@@ -104,6 +103,22 @@ Show results and fix any issues.
 
 ### 5. Tests
 
+Resolve the test framework via the
+[`toolchain-resolver`](../../contexts/execution/toolchain-resolver.md) before
+writing anything — write the regression test in the framework the project
+actually uses, never a hard-coded one.
+
+**The resolver is the only stack authority, and `skills:` deliberately names
+no single-stack skill.** This command's 195 lines of body mention no framework
+at all, so the `pest-testing` binding and the `framework: laravel` marker it
+used to carry made a Python or Go repository load PHP testing guidance for a
+command that never asks for it. The load is **resolver-conditional**: once the
+runner is named, load the skill for THAT runner if one exists —
+`pest-testing` for pest, `laravel` for `php artisan test`,
+`playwright-testing` for playwright. Unconditional became conditional; it did
+not become nothing. Same contract as `/tests create` and `/tests execute`,
+which were de-bound first.
+
 Check for existing tests that cover the affected code:
 
 ```

package/dist/agent-src/commands/bug/investigate.md

@@ -26,7 +26,7 @@ packs:
 
 Run `git branch --show-current` and extract ticket IDs:
 
-- Match pattern: `[A-Z]+-[0-9]+` (e.g., `DEV-1234`, `PROJ-567`)
+- Read it with `agent-config git:convention ticket` ([`commit-subject`](../../../../skills/git-workflow/references/commit-subject.md) § Reading the ticket; e.g. `DEV-1234`, `PROJ-567`)
 - Common branch formats: `fix/DEV-1234/description`, `hotfix/DEV-999`, `fix/bug-description`
 
 If a ticket ID is found:

package/dist/agent-src/commands/commit/in-chunks.md

@@ -43,10 +43,16 @@ If there are no uncommitted changes (staged or unstaged), report
 
 ### 2. Determine the ticket number
 
-- Extract the ticket ID from the current branch name (e.g. `feat/DEV-1234/...`
-  → `DEV-1234`).
-- If no ticket ID is found, omit the scope from the messages — write
-  `chore: ...` not `chore(): ...`. Do **not** ask the user for one.
+- Read the ticket exactly as [`/commit`](../commit.md) step 2 does
+  (`agent-config git:convention ticket`, with the card's `ticket_keys` line as
+  `--keys`) and place it per the convention in force —
+  [`commit-subject`](../../../../skills/git-workflow/references/commit-subject.md),
+  including its area-scope clause. A candidate marked `unknown-key` is omitted,
+  never asked about.
+- If no ticket ID is found, omit it — write `chore: ...` not `chore(): ...`.
+  Do **not** ask the user for one.
+  Under an approved convention card this is never a stop: the family's form
+  without the ticket part is valid (`commit-subject` § Placing the ticket).
 
 ### 3. Analyze and split
 
@@ -69,6 +75,12 @@ Splitting rules:
 
 Generate commit messages per [`commit-conventions`](../../rules/commit-conventions.md).
 
+**No convention established — never ask.** When `agent-config git:convention
+show --key commit_format` prints `no convention established`, run
+`agent-config git:convention measure`, carry its `verdict` line into the summary,
+and generate under the convention in force (the default). Asking and writing the
+card belong to the interactive [`/commit`](../commit.md) step 3c only.
+
 ### 3b. Secret-leak pre-flight — MANDATORY, not autonomy-lifted
 
 Run `./scripts-run src/scripts/check_secret_leak`. A high-confidence hit is a
@@ -79,14 +91,22 @@ strip one.
 
 ### 4. Validate, then commit immediately
 
-Before staging anything, run the same `preview-on-error` validator as
-[`/commit`](../commit.md) step 5. Each generated message must match
-`^(feat|fix|chore|docs|refactor|test|perf|style|build|ci|revert)(\([^)]+\))?!?: .+`.
+Before staging anything, pipe every generated subject through
+`agent-config git:convention subject`, as [`/commit`](../commit.md) step 5 does.
+
+- **Exit `0`** → proceed silently; a `note:` line naming a `commit-msg` hook
+  or a commitlint config is not a stop — that validator runs at commit, and a
+  rejection there stops the run like any failed commit.
+- **Exit `1`** → stop, print the verb's output (each failed subject and its
+  rule, or the format it could not read), hand back to the user.
+  Do **not** auto-commit broken messages even though confirmation is
+  otherwise suppressed.
 
-- **All valid** → proceed silently.
-- **Any invalid** → stop, print the failed message(s) + the regex,
-  hand back to the user. Do **not** auto-commit broken messages even
-  though confirmation is otherwise suppressed.
+An approved convention card an earlier [`/commit`](../commit.md) wrote and left
+uncommitted (`agents/memory/curated/conventions/approved/commit-subject.md`,
+unstaged or untracked) rides in the first chunk: add it with that chunk's files
+(`git add -- agents/memory/curated/conventions/approved/commit-subject.md`), so
+every worktree and clone reads the convention it was approved under.
 
 For each planned commit in order:
 
@@ -98,29 +118,18 @@ git commit -m "{message}"
 No "looks good?" prompt. No confirmation step. The user invoked this
 command knowing the plan would be executed.
 
-### 5. Report (verbosity-gated)
-
-Read `verbosity.post_action_reports` from `.agent-settings.yml` (default
-`minimal`). Same contract as [`/commit`](../commit.md) step 7.
-
-- `off` → emit nothing (still run `git log --oneline -N` silently for
-  the agent's own verification).
-- `minimal` (default) → one line:
+### 5. Report
 
-  ```
-  → N commits created
-  ```
+Same contract as [`/commit`](../commit.md) step 7 — one line (run
+`git log --oneline -N` silently for the agent's own verification):
 
-- `full` → full summary:
-
-  ```
-  Created N commits:
-  1. {sha1}  feat(DEV-1234): {summary}
-  2. {sha2}  test(DEV-1234): {summary}
-  3. {sha3}  chore: {summary}
-  ```
+```
+→ N commits created
+```
 
-  Include `git log --oneline -N` output for verification.
+A run that measured because no convention was
+established adds the verb's `verdict` line and `→ run /commit once to choose and
+record the convention`.
 
 ## Rules
 

package/dist/agent-src/commands/commit.md

@@ -57,7 +57,13 @@ If there are no uncommitted changes (staged or unstaged), report "Nothing to com
 
 ### 2. Determine the ticket number
 
-- Extract the ticket ID from the current branch name (e.g. `feat/DEV-1234/...` → `DEV-1234`).
+- Read the ticket from the current branch name:
+  `agent-config git:convention ticket --keys "<the ticket_keys line of the approved convention card>"`
+  (no card → no `--keys`). The first line is `ticket <ID>` or `ticket none`
+  (`feat/DEV-1234/...`, `fix/DEV-1234-quantity` → `DEV-1234`; `fix/CVE-2026-12345-patch` → none).
+  A candidate marked `unknown-key` → ask the user whether it is a ticket; a yes
+  adds its key to the card's `ticket_keys`. A `proposal` line → offer it for the
+  card, never write it unasked.
 - If no ticket ID is found in the branch name, ask the user:
   ```
   > No Jira ticket found in branch name. Do you want to include one?
@@ -65,8 +71,9 @@ If there are no uncommitted changes (staged or unstaged), report "Nothing to com
   > 1. Yes — I'll provide the ticket number
   > 2. No — skip ticket number
   ```
-- If the user provides a ticket number, use it as the scope in all commit messages.
-- If skipped, omit the scope entirely — write `chore: ...` not `chore(): ...`.
+- Place the ticket (given or read) in every message per the convention in force —
+  [`commit-subject`](../../../skills/git-workflow/references/commit-subject.md) § Placing the ticket;
+  skipped → omit it, never `chore(): …`.
 
 ### 3. Analyze the changes
 
@@ -95,6 +102,43 @@ once the user resolves it (move to a store, or add an audited
 `# secret-allow` / `.secret-allow` entry for a confirmed false positive). Never
 silently commit a secret; never silently strip one.
 
+### 3c. Establish the convention — once per repository
+
+Before any subject is generated, run
+`agent-config git:convention show --key commit_format`. It prints
+`no convention established — run git:convention measure` only when there is
+no declaration, no approved card and no commitlint config; otherwise skip this step.
+
+1. Run `agent-config git:convention measure`. Its `verdict` line is the input to
+   the one question below; the thresholds behind it are the `measure` rows of
+   [`commit-subject`](../../../skills/git-workflow/references/commit-subject.md) § The grammar.
+2. Ask **once**, as numbered options, naming the family, its share, the capped
+   total and the author count from the verb's output:
+   - `verdict established: <family>` → the established family (recommended) vs
+     Conventional Commits;
+   - `verdict below the bar` → the two strongest families (each with its share)
+     vs Conventional Commits, quoting the verb's reasons.
+3. Run `agent-config git:convention measure --family <answer>` (Conventional
+   Commits is `--family conventional`) and write its `card` block to
+   `agents/memory/curated/conventions/approved/commit-subject.md` — **also when
+   the answer is Conventional Commits**, so this question is never asked again in
+   this repository. The card is the one file this command writes. Stage it
+   explicitly with `git add -- agents/memory/curated/conventions/approved/commit-subject.md`
+   and commit it in the first commit of this run — the commit it was approved
+   for — so a worktree or a fresh clone reads the same answer instead of asking
+   again; name it in the report.
+4. When the same output carries a `team file` block, show it: it is the
+   ready-to-commit `.git-convention.yml` (`commit_format`, `branch_pattern`,
+   never `update_strategy`). Ask once whether to create it. Only on the user's
+   explicit yes this turn run `agent-config git:convention init --yes` with the
+   shown values (`--commit-format`, `--branch-pattern`); it creates the file once,
+   refuses when one exists and never commits it — name it in the report. On no,
+   print the content for a human to create and commit. Never write the file by
+   hand, and never change an existing one.
+
+Then generate every subject under the answer: `git:convention subject` reads the
+card the moment it exists.
+
 ### 4. Plan the commits
 
 For each logical group, determine the commit message following the commit conventions rule
@@ -107,40 +151,42 @@ Rules for splitting:
 - **Style-only changes** (formatter / auto-refactor output — ECS, Prettier, Ruff) may get their own `style:` or `chore:` commit
   if they are large and mixed with logic changes.
 
-### 5. Present the commit plan (verbosity-gated)
-
-Read `verbosity.preview_artifacts` and `verbosity.routine_confirmations`
-from `.agent-settings.yml`. Both default to `false`.
+### 5. Present the commit plan
 
-**Terse path** — `preview_artifacts: false` AND `routine_confirmations: false`:
+**Terse path** — the default; there is no setting that turns it off:
 
-1. Validate every generated commit message against the
-   conventional-commits regex
-   `^(feat|fix|chore|docs|refactor|test|perf|style|build|ci|revert)(\([^)]+\))?!?: .+`.
+1. Pipe every generated subject, one per line, through
+   `agent-config git:convention subject` — it resolves the convention in force
+   and validates against it. Act on the exit code: `0` valid · `1` the output
+   lists each failure and its rule → treat as invalid. A `note:` line naming a
+   `commit-msg` hook or a commitlint config means that validator also runs at
+   commit and may be stricter; it never changes the exit.
 2. **All messages valid** → skip the preview block and the confirmation
    prompt. Print one line summarising the plan and proceed to step 6:
 
    ```
-   → 3 commits planned: feat, test, chore (scope: DEV-1234)
+   → 3 commits planned: feat, test, chore (ticket: DEV-1234)
    ```
 
 3. **Any message invalid** → `preview-on-error` safety net fires:
-   force the full preview block below + the numbered confirm prompt,
-   regardless of the two flags. The user must approve before step 6.
+   force the full preview block below + the numbered confirm prompt.
+   The user must approve before step 6.
 4. **Hard-Floor diff** (bulk deletion ≥5 unrelated files, infra changes
    touching Terraform / Pulumi / k8s / Ansible / cloud-config) →
-   ALWAYS preview + confirm regardless of flags, per
+   ALWAYS preview + confirm, per
    [`non-destructive-by-default`](../rules/non-destructive-by-default.md).
 
-**Preview path** — `preview_artifacts: true` OR `routine_confirmations: true`
-(or `preview-on-error` triggered):
+**Preview path** — `preview-on-error` or a Hard-Floor diff triggered it, or the
+user asked for the preview (override below):
 
 Show the proposed commits as a numbered list, including which files go into each:
 
 ```
 Proposed commits:
 
-(Laravel-project example)
+(Laravel-project example, `git.commit_format: ticket-scope`; under
+`ticket-conventional` the first subject reads
+`DEV-1234 feat(working-time): add absence type filter to working time report`)
 1. feat(DEV-1234): add absence type filter to working time report
    → app/Services/WorkingTimeService.php
    → app/Http/Controllers/WorkingTimeController.php
@@ -163,8 +209,8 @@ Then ask:
 Wait for the user's response before doing anything.
 
 **Override:** the user may force the preview at any time with
-*"show me the commit plan first"* / *"preview commits"* — treat as a
-one-shot `preview_artifacts: true` for this invocation.
+*"show me the commit plan first"* / *"preview commits"* — take the preview
+path for this invocation.
 
 ### 6. Commit
 
@@ -175,26 +221,19 @@ For each planned commit in order:
 1. Stage only the files for that commit: `git add {files...}`
 2. Commit: `git commit -m "{message}"`
 
-### 7. Report (verbosity-gated)
-
-Read `verbosity.post_action_reports` from `.agent-settings.yml` (default
-`minimal`).
-
-- `off` → emit nothing.
-- `minimal` (default) → one line:
+### 7. Report
 
-  ```
-  → 3 commits created
-  ```
+One line:
 
-- `full` → multi-line summary: number of commits + commit messages
-  (one per line).
+```
+→ 3 commits created
+```
 
 ## Rules
 
 - **Preview path: never commit before the user confirms the plan — once, for the whole plan.** On the terse path (`preview_artifacts: false` AND `routine_confirmations: false`) the `/commit` invocation itself is the confirmation (`commit-policy` exception 3); proceed without a prompt.
 - **Never push** — pushing is left to the user.
-- **Never modify files** — only stage and commit existing changes.
+- **Never modify files** — only stage and commit existing changes. The one exception is the convention card step 3c writes after the user's answer, committed with the first commit of the run.
 - **Do NOT add untracked files** unless they are clearly part of the change (check with `git status`).
 - **Follow commit conventions** as defined in `.augment/rules/commit-conventions.md`.
 

package/dist/agent-src/commands/condense.md

@@ -89,15 +89,9 @@ Must show ✅ (`dist/agent-src/` in sync with source).
 Must show **zero 🔴 errors**. Warnings (🟡) are acceptable.
 If any 🔴 errors remain: go back and fix those files before finishing.
 
-## Step 5: Summary (verbosity-gated)
+## Step 5: Summary
 
-Read `verbosity.post_action_reports` from `.agent-settings.yml` (default
-`minimal`).
-
-- `off` → emit nothing on success; surface errors only.
-- `minimal` (default) → one line: `→ N files synced`.
-- `full` → multi-line table with per-category stats (files synced,
-  stale files reaped).
+One line on success: `→ N files synced`. Errors are always surfaced.
 
 ## Iron Laws — preserved by construction
 

package/dist/agent-src/commands/council/debate.md

@@ -67,20 +67,24 @@ this automatically when invoked via `debate`). The estimate covers
 projection. Surface the projection to the user with a clear note that
 progressive disclosure may stop the debate early.
 
-### 4. Spend bound (ask when a billable member has no ceiling)
+### 4. Spend projection (information, never a prompt)
 
-Per [`ai-council` skill § Procedure 3](../../skills/ai-council/SKILL.md)
-and ADR-230, surface the projected total spend in every case, and ask
-only when a billable member is in the fan-out AND both
-`cost_budget.max_total_usd` and `daily_limit_usd` are `0`.
+Per [`ai-council` skill § Procedure 3](../../skills/ai-council/SKILL.md),
+ADR-230 and ADR-279, surface the projected total spend in every case and
+**do not ask** because no cap is set. A cap nobody configured is not a
+bound, and its absence is not a reason to ask.
 
 A debate multiplies the per-round cost by `N`, so quote the worst case,
 not one round — that is what a ceiling has to cover. The between-rounds
 confirmation gate is unaffected: it is a separate, per-round control and
 it still fires, which is where a debate that turns out expensive gets
-stopped.
+stopped. `debate.max_cost_usd` likewise still refuses above whatever
+figure someone set; it ships at `0`, i.e. no refusal cap.
 
-Numbered options, when the ask applies:
+The round shape is still the user's choice, and it is not a spend
+question — it decides whether the between-round confirmation fires,
+which this change does not touch. Offer it regardless of whether a cap
+is set:
 
 > The debate is projected to cost **$X.XXXX** ($Y.YYYY × N rounds, worst case).
 >

package/dist/agent-src/commands/council/default.md

@@ -151,22 +151,23 @@ Construct each member from the resolved mode:
 - `manual` → `ManualClient` from `scripts.ai_council.clients`
   (`billable=False`, no API key, no SDK call).
 
-### 3. Spend bound — ASK when a billable member has no ceiling
+### 3. Spend estimate — information, never a prompt
 
-Council calls to billable members spend money, so a bound is mandatory.
-Per ADR-230 the bound is a **ceiling**, not a per-invocation approval —
-the same three cases the `ai-council` skill resolves in its Procedure § 3:
+Per ADR-230 the bound is a **ceiling**, not a per-invocation approval,
+and per ADR-279 a ceiling applies only where someone set one. The three
+cases the `ai-council` skill resolves in its Procedure § 3 now agree on
+the ask and differ only in what a cap does once it exists:
 
 | Situation | Behaviour |
 |---|---|
 | No billable member (all `mode: cli`, subscription auth) | No gate. Estimate is information; go to Step 4. Spend is $0. |
 | Billable member **with** `cost_budget.max_total_usd` or `daily_limit_usd` non-zero | No per-run ask. The ceiling is the authorization the user already gave; `on_overrun` still pauses per member on breach. |
-| Billable member with **both** caps `0` | **Ask.** Nothing bounds the spend, so the user must. |
+| Billable member with **both** caps `0` — the shipped default | No ask. Render the estimate and go to Step 4; no cap exists, so nothing can breach. |
 
-`personal.autonomy` neither creates nor lifts that bound — autonomy is
-not a ceiling. Consumers who want a per-invocation gate set a small
-`cost_budget.max_total_usd`; every call then breaches and `on_overrun`
-asks per member.
+`personal.autonomy` neither creates nor lifts a bound — autonomy is
+not a ceiling, and neither is its absence. Consumers who want a
+per-invocation gate set a small `cost_budget.max_total_usd`; every call
+then breaches and `on_overrun` asks per member.
 
 Run the CLI in **estimate** mode first — it bundles the artefact, runs
 redaction, and prints the per-member preview without spending:

package/dist/agent-src/commands/council.md

@@ -34,7 +34,7 @@ sub-command swaps the mode-specific addendum.
 
 1. Resolve the target + capture the original ask.
 2. Check the council is configured + price table fresh.
-3. Spend bound (ask only when a billable member has no ceiling — ADR-230).
+3. Spend estimate — rendered as information, never a prompt (ADR-230, ADR-279).
 4. Run the CLI.
 5. Render the report (5 / 5a / 5b — render → critical lens → user options).
 6. Hard floor — text only.

package/dist/agent-src/commands/estimate-ticket.md

@@ -28,7 +28,7 @@ packs:
 Accept the same four input paths as [`/refine-ticket`](refine-ticket.md):
 
 1. **Explicit key** — `/estimate-ticket PROJ-123`
-2. **Branch detection** — no arg → `git branch --show-current` + regex `[A-Z]+-[0-9]+`
+2. **Branch detection** — no arg → `agent-config git:convention ticket` on the current branch ([`commit-subject`](../../../skills/git-workflow/references/commit-subject.md) § Reading the ticket)
 3. **Pasted text** — markdown block under the command
 4. **URL** — `/estimate-ticket https://…/browse/PROJ-123`
 

package/dist/agent-src/commands/feature/explore.md

@@ -38,8 +38,8 @@ Before entering brainstorm mode, check whether the input looks like an
 **existing ticket** rather than a fresh idea. If any of the following
 match, surface the hint below and let the user pick:
 
-- A Jira/Linear URL is present (`https://.../browse/[A-Z]+-[0-9]+`)
-- The current branch name contains a ticket key (`[A-Z]+-[0-9]+`)
+- A Jira/Linear URL is present (`https://.../browse/<ticket key>`)
+- The current branch name contains a ticket key (`agent-config git:convention ticket` prints one)
 - The pasted text contains structured AC bullets (`- [ ]`, "Acceptance Criteria", "AC:")
 
 Prompt:
@@ -56,7 +56,7 @@ continue with Step 2.
 ### 2. Gather external context
 
 **Auto-detect ticket from branch:**
-Run `git branch --show-current` and extract ticket IDs (pattern: `[A-Z]+-[0-9]+`).
+Read the ticket ID from the current branch with `agent-config git:convention ticket` ([`commit-subject`](../../../../skills/git-workflow/references/commit-subject.md) § Reading the ticket).
 
 If a ticket ID is found:
 ```

package/dist/agent-src/commands/feature/plan.md

@@ -53,7 +53,7 @@ Then, once that is answered:
 ### 2. Gather external context
 
 **Auto-detect ticket from branch:**
-Run `git branch --show-current` and extract ticket IDs (pattern: `[A-Z]+-[0-9]+`).
+Read the ticket ID from the current branch with `agent-config git:convention ticket` ([`commit-subject`](../../../../skills/git-workflow/references/commit-subject.md) § Reading the ticket).
 
 If a ticket ID is found:
 ```
@@ -402,7 +402,7 @@ Declared by `produces_roadmap: true` in the frontmatter and enforced by
 `lint_roadmap_producers`: a producer that does not end here reds CI.
 
 Read `planning.closure_pass` (missing = `true`). When active, run
-[`/challenge-me closure`](../../../meta/challenge-me/closure/command.md) on the
+`/challenge-me closure` on the
 roadmap this command just produced, before handing back:
 
 ```bash