kaizen-agent 1.7.1 → 1.7.2

package/skills/kaizen/adapters.md

@@ -1,77 +1,77 @@
-# Adapters — running kaizen outside Claude Code
-
-`spec.md` plus its per-stage files (`spec-plan.md`, `spec-build.md`, `spec-review.md`,
-`spec-main.md`) are the workflow. It does not mention any vendor, and it is the single
-source of truth. An adapter is a thin file that points a given tool at it and
-describes how that tool dispatches stages. Adapters never restate the stage rules;
-duplicating them is how the tools drift apart.
-
-The portability comes from `.kaizen/` being plain files in the repository. Any tool
-that can read a directory can resume a run another tool started.
-
-## Claude Code
-
-Native. The skill in this directory dispatches each stage through the Agent tool
-using the `kaizen-planner`, `kaizen-builder`, and `kaizen-reviewer` agent
-definitions, and the main thread owns `state.json`.
-
-## Codex — `AGENTS.md`
-
-Codex reads `AGENTS.md` from the repository root. Append:
-
-```markdown
-## Kaizen workflow
-
-Non-trivial work in this repo follows the staged workflow in
-`.kaizen/spec.md`: plan, human approval, implement, independent review, bounded
-fix loop. Read that file before starting, and read the config chain (`.kaizen/config.yml`, `~/.kaizen/config.yml`, `config.default.yml`) for the
-mode and approval settings.
-
-Run state lives in `.kaizen/runs/<id>/`. Always read `state.json` first and
-continue from the stage it names. Never advance a run whose `awaiting` field is
-non-null without a recorded human decision in `02-approval.md` — another tool or
-session may be mid-run.
-
-Codex has no separate subagents, so run the stages sequentially in one session,
-but keep them clean: start each stage by reading only that stage's declared inputs
-from its stage file, and write its artifact before moving on. When reviewing, judge the
-diff on its own terms rather than defending the reasoning you used while writing it.
-```
-
-Copy `spec*.md` into `.kaizen/` so the reference resolves
-without this skill installed.
-
-## Antigravity
-
-Antigravity reads `AGENTS.md` as well; use the same block. If a workspace-level rules
-file is configured instead, point it at `.kaizen/spec.md` and the stage file being run with the same three
-paragraphs.
-
-Where Antigravity can run parallel agents, map planner, builder, and reviewer onto
-separate agents exactly as Claude Code does — the independent reviewer context is the
-part worth preserving in any port.
-
-## Handing a run between tools
-
-Nothing special is required. Because every stage's input and output are files:
-
-1. Tool A finishes a stage and updates `state.json`.
-2. Tool B opens the repo, reads `state.json`, sees `stage: "review"` and
-   `awaiting: null`, and runs the reviewer.
-
-The only rule is that whichever tool holds the run updates `state.json` when a stage
-completes. A stale `state.json` is the one way this breaks.
-
-## Installing the adapter files
-
-`/kaizen init` writes `.kaizen/` with the `spec*.md` files into the
-project folder, plus the gitignore entry when that folder is a git repository.
-`/kaizen install` additionally offers to append the `AGENTS.md` block above.
-`.kaizen/spec*.md` and `.kaizen/config.yml` should be committed even though run state
-is ignored — the workflow is shared, the run data is not:
-
-```gitignore
-.kaizen/*
-!.kaizen/spec*.md
-!.kaizen/config.yml
-```
+# Adapters — running kaizen outside Claude Code
+
+`spec.md` plus its per-stage files (`spec-plan.md`, `spec-build.md`, `spec-review.md`,
+`spec-main.md`) are the workflow. It does not mention any vendor, and it is the single
+source of truth. An adapter is a thin file that points a given tool at it and
+describes how that tool dispatches stages. Adapters never restate the stage rules;
+duplicating them is how the tools drift apart.
+
+The portability comes from `.kaizen/` being plain files in the repository. Any tool
+that can read a directory can resume a run another tool started.
+
+## Claude Code
+
+Native. The skill in this directory dispatches each stage through the Agent tool
+using the `kaizen-planner`, `kaizen-builder`, and `kaizen-reviewer` agent
+definitions, and the main thread owns `state.json`.
+
+## Codex — `AGENTS.md`
+
+Codex reads `AGENTS.md` from the repository root. Append:
+
+```markdown
+## Kaizen workflow
+
+Non-trivial work in this repo follows the staged workflow in
+`.kaizen/spec.md`: plan, human approval, implement, independent review, bounded
+fix loop. Read that file before starting, and read the config chain (`.kaizen/config.yml`, `~/.kaizen/config.yml`, `config.default.yml`) for the
+mode and approval settings.
+
+Run state lives in `.kaizen/runs/<id>/`. Always read `state.json` first and
+continue from the stage it names. Never advance a run whose `awaiting` field is
+non-null without a recorded human decision in `02-approval.md` — another tool or
+session may be mid-run.
+
+Codex has no separate subagents, so run the stages sequentially in one session,
+but keep them clean: start each stage by reading only that stage's declared inputs
+from its stage file, and write its artifact before moving on. When reviewing, judge the
+diff on its own terms rather than defending the reasoning you used while writing it.
+```
+
+Copy `spec*.md` into `.kaizen/` so the reference resolves
+without this skill installed.
+
+## Antigravity
+
+Antigravity reads `AGENTS.md` as well; use the same block. If a workspace-level rules
+file is configured instead, point it at `.kaizen/spec.md` and the stage file being run with the same three
+paragraphs.
+
+Where Antigravity can run parallel agents, map planner, builder, and reviewer onto
+separate agents exactly as Claude Code does — the independent reviewer context is the
+part worth preserving in any port.
+
+## Handing a run between tools
+
+Nothing special is required. Because every stage's input and output are files:
+
+1. Tool A finishes a stage and updates `state.json`.
+2. Tool B opens the repo, reads `state.json`, sees `stage: "review"` and
+   `awaiting: null`, and runs the reviewer.
+
+The only rule is that whichever tool holds the run updates `state.json` when a stage
+completes. A stale `state.json` is the one way this breaks.
+
+## Installing the adapter files
+
+`/kaizen init` writes `.kaizen/` with the `spec*.md` files into the
+project folder, plus the gitignore entry when that folder is a git repository.
+`/kaizen install` additionally offers to append the `AGENTS.md` block above.
+`.kaizen/spec*.md` and `.kaizen/config.yml` should be committed even though run state
+is ignored — the workflow is shared, the run data is not:
+
+```gitignore
+.kaizen/*
+!.kaizen/spec*.md
+!.kaizen/config.yml
+```

package/skills/kaizen/config.default.yml

@@ -1,167 +1,167 @@
-# Kaizen defaults. Settings live in ~/.kaizen/config.yml and apply to every project.
-# To make one project different, copy this to .kaizen/config.yml there and override
-# only the keys you care about; anything absent falls back to the global file, then
-# to this one.
-
-# Quick preset: low | medium | ultra | custom
-# Selecting low, medium, or ultra applies its bundled defaults for runner, executor,
-# approvals, and auto-fix. Modifying individual keys sets this to custom.
-preset: medium
-
-# plan-only | approve | auto | review-only
-mode: approve
-
-approvals:
-  # Stop after the plan, before any code is written.
-  plan: true
-  # Stop after the review, before the work is called done. Off by default: when
-  # auto_fix has already closed the findings at or above its threshold, this second
-  # stop mostly asks the user to approve a report they are about to read anyway.
-  review: false
-  # Confirm every individual file edit. Maximum control, slowest.
-  each_file: false
-
-# How much machinery a run uses.
-#   full - each stage is its own agent, started cold. The reviewer never sees how the
-#          work was done, which is the only reason its findings are worth reading.
-#          Costs several times what the same change costs in one session.
-#   lite - the session you are talking to runs every stage itself, in order, keeping
-#          each stage to its own declared inputs. Roughly the cost of doing the work
-#          by hand, and you watch it happen. The reviewer has seen the work, so
-#          expect it to miss what it just did.
-# Everything else is the same either way: the plan, the approval, the gate, the
-# findings, the files, and resuming a run in another tool.
-# Override for one run with `/kaizen lite <request>` or `/kaizen full <request>`.
-runner: full
-
-build:
-  # Who carries out the approved plan.
-  #   subagent - dispatch the builder as its own agent. The main thread keeps its
-  #              context and the work is reported back. The default, and the only
-  #              sane option for a large run.
-  #   inline   - the main thread does the work itself, against the same plan and the
-  #              same contract. You watch it happen and can interrupt mid-step, at
-  #              the cost of main-thread context.
-  #   ask      - offer the choice at the plan approval.
-  # The reviewer is a separate agent in every case; that is not configurable, because
-  # an agent reviewing work it just did will defend it.
-  executor: subagent
-
-auto_fix:
-  # false: every finding stops the run and becomes a user decision.
-  # true:  findings at or above min_severity are fixed and re-reviewed automatically.
-  enabled: true
-  # critical | high | medium | low
-  min_severity: high
-  # Fix/recheck rounds before remaining findings are escalated to the user. Each round
-  # is a full builder pass plus a full reviewer pass, so this is the most expensive
-  # knob in the file. A third round rarely closes what two could not.
-  max_iterations: 2
-
-track:
-  # The kind of work a run is doing, which decides what "deliverable",
-  # "verification", and "finding" mean for every stage. It is INFERRED from the
-  # request by the planner and stated back in the plan's Framing for the user to
-  # correct at the approval — the user is never asked to categorize their own request.
-  # Set this only to pin a default for a repo that is always one kind of work
-  # (e.g. a docs repo). auto (or empty) infers per run.
-  default: auto
-
-review:
-  # Checks that run in every track.
-  checks:
-    - correctness
-    - plan_conformance
-  # Additional checks per track, chosen by what the run actually is. The planner may
-  # name a track not listed here; the reviewer then derives its checks from the
-  # track's "what would make it wrong".
-  # Each check below is one question the reviewer asks; the line beside it says which.
-  track_checks:
-    # security - no new way in, no secret or user data exposed
-    # reuse_simplification - nothing rebuilt that the codebase already had
-    # test_coverage - the changed behaviour has a check that fails when it breaks
-    software: [security, reuse_simplification, test_coverage]
-    # blast_radius and regression are universal — see spec-review.md — not listed here.
-    # completeness - every part the piece promises is actually present
-    # claim_accuracy - every checkable claim traced to a source
-    # reader_usability - its intended reader can act on it without asking
-    writing: [completeness, claim_accuracy, reader_usability]
-    # audience_fit - written for the people who will actually read it
-    # voice_consistency - reads as the same voice as the surrounding material
-    communication: [claim_accuracy, audience_fit, voice_consistency]
-    # prerequisites - everything needed before step one is stated
-    # rollback - a way back exists for every step that changes something
-    # executability - someone else can run it start to finish unaided
-    operations: [prerequisites, rollback, executability]
-    # evidence_support - the conclusion follows from evidence shown, not asserted
-    # selection_bias - the evidence was not picked to fit the answer
-    # alternatives - the readings that were rejected are named and why
-    research: [evidence_support, selection_bias, alternatives]
-  # Let the reviewer append durable lessons to .kaizen/memory.md.
-  write_memory: true
-  # Every stage of every future run reads memory.md, so cap it. Over this, the reviewer
-  # prunes stale or superseded lessons before appending rather than growing the file.
-  memory_max_lines: 100
-
-backlog:
-  # Capture what a run deliberately did not do: out-of-scope suggestions the user did
-  # not take, findings below the fix threshold, and debts deferred at an approval.
-  # Written by the main thread at the plan approval and again at the final approval.
-  enabled: true
-  # Per run, beside its plan and review — provenance without a central file to sync.
-  file: 06-backlog.md
-  # Items are open | done | rejected. Only open ones are printed by /kaizen backlog;
-  # closed ones stay in their run's file as record.
-  statuses: [open, done, rejected]
-  # Before state.keep_runs prunes a run, its still-open items move here, relative to
-  # state.dir. Done and rejected items are pruned with the run.
-  orphan_file: backlog.md
-
-state:
-  dir: .kaizen
-  # Keep this many completed runs before pruning the oldest. Pruning first rescues
-  # open backlog items into backlog.orphan_file, so it never destroys open work.
-  keep_runs: 20
-
-git:
-  # Never commit, push, or open a PR unless the user explicitly asks.
-  # Ignored when the project is not a git repository.
-  auto_commit: false
-  # Create a branch before the implement stage when on the default branch.
-  # Ignored when the project is not a git repository.
-  branch_before_implement: true
-
-ui:
-  # Mouse in the TUI: click a row or card to select it, click it again to open or
-  # run it, wheel to scroll. Off by default, and the reason is a real trade: while
-  # the TUI holds the mouse, the terminal cannot select text with it, so copying a
-  # run id out of the dashboard stops working until you leave.
-  mouse: false
-
-subagents:
-  # Claude Code only: other tools (Codex, Antigravity, opencode, gemini) have no
-  # per-stage model knob and run every stage on whatever the tool itself is set to.
-  # Model each stage subagent runs on, when the runner is full.
-  #   inherit - whatever the main session uses. The default.
-  #   opus | sonnet | haiku - pin it. A cheaper model for the builder is the usual
-  #   reason to touch this; the reviewer on a cheaper model finds less.
-  model: inherit
-  # Reasoning effort for the stage subagents: inherit | low | medium | high.
-  # Set through the agent definition's frontmatter.
-  effort: inherit
-
-agent:
-  # Default coding agent to launch from dashboard backlog (auto | claude | codex | agy | opencode | gemini).
-  # auto picks Claude Code if installed, otherwise the first agent found on PATH.
-  default: auto
-  # Per-tool model and effort for the session the board launches, passed as that
-  # tool's own flags (claude --model/--effort, agy --model/--effort, codex -m and
-  # model_reasoning_effort, gemini -m, opencode -m). Model names are the tool's own:
-  # `agy models`, `opencode models`, Claude's opus|sonnet|haiku. Absent = tool default.
-  # Uncomment the tool you launch:
-  # claude: { model: opus, effort: high }
-  # codex: { model: gpt-5, effort: low }
-  # agy: { model: gemini-3.1-pro-low }
-  # opencode: { model: anthropic/claude-sonnet-4-6 }
-  # gemini: { model: gemini-2.5-pro }
-
+# Kaizen defaults. Settings live in ~/.kaizen/config.yml and apply to every project.
+# To make one project different, copy this to .kaizen/config.yml there and override
+# only the keys you care about; anything absent falls back to the global file, then
+# to this one.
+
+# Quick preset: low | medium | ultra | custom
+# Selecting low, medium, or ultra applies its bundled defaults for runner, executor,
+# approvals, and auto-fix. Modifying individual keys sets this to custom.
+preset: medium
+
+# plan-only | approve | auto | review-only
+mode: approve
+
+approvals:
+  # Stop after the plan, before any code is written.
+  plan: true
+  # Stop after the review, before the work is called done. Off by default: when
+  # auto_fix has already closed the findings at or above its threshold, this second
+  # stop mostly asks the user to approve a report they are about to read anyway.
+  review: false
+  # Confirm every individual file edit. Maximum control, slowest.
+  each_file: false
+
+# How much machinery a run uses.
+#   full - each stage is its own agent, started cold. The reviewer never sees how the
+#          work was done, which is the only reason its findings are worth reading.
+#          Costs several times what the same change costs in one session.
+#   lite - the session you are talking to runs every stage itself, in order, keeping
+#          each stage to its own declared inputs. Roughly the cost of doing the work
+#          by hand, and you watch it happen. The reviewer has seen the work, so
+#          expect it to miss what it just did.
+# Everything else is the same either way: the plan, the approval, the gate, the
+# findings, the files, and resuming a run in another tool.
+# Override for one run with `/kaizen lite <request>` or `/kaizen full <request>`.
+runner: full
+
+build:
+  # Who carries out the approved plan.
+  #   subagent - dispatch the builder as its own agent. The main thread keeps its
+  #              context and the work is reported back. The default, and the only
+  #              sane option for a large run.
+  #   inline   - the main thread does the work itself, against the same plan and the
+  #              same contract. You watch it happen and can interrupt mid-step, at
+  #              the cost of main-thread context.
+  #   ask      - offer the choice at the plan approval.
+  # The reviewer is a separate agent in every case; that is not configurable, because
+  # an agent reviewing work it just did will defend it.
+  executor: subagent
+
+auto_fix:
+  # false: every finding stops the run and becomes a user decision.
+  # true:  findings at or above min_severity are fixed and re-reviewed automatically.
+  enabled: true
+  # critical | high | medium | low
+  min_severity: high
+  # Fix/recheck rounds before remaining findings are escalated to the user. Each round
+  # is a full builder pass plus a full reviewer pass, so this is the most expensive
+  # knob in the file. A third round rarely closes what two could not.
+  max_iterations: 2
+
+track:
+  # The kind of work a run is doing, which decides what "deliverable",
+  # "verification", and "finding" mean for every stage. It is INFERRED from the
+  # request by the planner and stated back in the plan's Framing for the user to
+  # correct at the approval — the user is never asked to categorize their own request.
+  # Set this only to pin a default for a repo that is always one kind of work
+  # (e.g. a docs repo). auto (or empty) infers per run.
+  default: auto
+
+review:
+  # Checks that run in every track.
+  checks:
+    - correctness
+    - plan_conformance
+  # Additional checks per track, chosen by what the run actually is. The planner may
+  # name a track not listed here; the reviewer then derives its checks from the
+  # track's "what would make it wrong".
+  # Each check below is one question the reviewer asks; the line beside it says which.
+  track_checks:
+    # security - no new way in, no secret or user data exposed
+    # reuse_simplification - nothing rebuilt that the codebase already had
+    # test_coverage - the changed behaviour has a check that fails when it breaks
+    software: [security, reuse_simplification, test_coverage]
+    # blast_radius and regression are universal — see spec-review.md — not listed here.
+    # completeness - every part the piece promises is actually present
+    # claim_accuracy - every checkable claim traced to a source
+    # reader_usability - its intended reader can act on it without asking
+    writing: [completeness, claim_accuracy, reader_usability]
+    # audience_fit - written for the people who will actually read it
+    # voice_consistency - reads as the same voice as the surrounding material
+    communication: [claim_accuracy, audience_fit, voice_consistency]
+    # prerequisites - everything needed before step one is stated
+    # rollback - a way back exists for every step that changes something
+    # executability - someone else can run it start to finish unaided
+    operations: [prerequisites, rollback, executability]
+    # evidence_support - the conclusion follows from evidence shown, not asserted
+    # selection_bias - the evidence was not picked to fit the answer
+    # alternatives - the readings that were rejected are named and why
+    research: [evidence_support, selection_bias, alternatives]
+  # Let the reviewer append durable lessons to .kaizen/memory.md.
+  write_memory: true
+  # Every stage of every future run reads memory.md, so cap it. Over this, the reviewer
+  # prunes stale or superseded lessons before appending rather than growing the file.
+  memory_max_lines: 100
+
+backlog:
+  # Capture what a run deliberately did not do: out-of-scope suggestions the user did
+  # not take, findings below the fix threshold, and debts deferred at an approval.
+  # Written by the main thread at the plan approval and again at the final approval.
+  enabled: true
+  # Per run, beside its plan and review — provenance without a central file to sync.
+  file: 06-backlog.md
+  # Items are open | done | rejected. Only open ones are printed by /kaizen backlog;
+  # closed ones stay in their run's file as record.
+  statuses: [open, done, rejected]
+  # Before state.keep_runs prunes a run, its still-open items move here, relative to
+  # state.dir. Done and rejected items are pruned with the run.
+  orphan_file: backlog.md
+
+state:
+  dir: .kaizen
+  # Keep this many completed runs before pruning the oldest. Pruning first rescues
+  # open backlog items into backlog.orphan_file, so it never destroys open work.
+  keep_runs: 20
+
+git:
+  # Never commit, push, or open a PR unless the user explicitly asks.
+  # Ignored when the project is not a git repository.
+  auto_commit: false
+  # Create a branch before the implement stage when on the default branch.
+  # Ignored when the project is not a git repository.
+  branch_before_implement: true
+
+ui:
+  # Mouse in the TUI: click a row or card to select it, click it again to open or
+  # run it, wheel to scroll. Off by default, and the reason is a real trade: while
+  # the TUI holds the mouse, the terminal cannot select text with it, so copying a
+  # run id out of the dashboard stops working until you leave.
+  mouse: false
+
+subagents:
+  # Claude Code only: other tools (Codex, Antigravity, opencode, gemini) have no
+  # per-stage model knob and run every stage on whatever the tool itself is set to.
+  # Model each stage subagent runs on, when the runner is full.
+  #   inherit - whatever the main session uses. The default.
+  #   opus | sonnet | haiku - pin it. A cheaper model for the builder is the usual
+  #   reason to touch this; the reviewer on a cheaper model finds less.
+  model: inherit
+  # Reasoning effort for the stage subagents: inherit | low | medium | high.
+  # Set through the agent definition's frontmatter.
+  effort: inherit
+
+agent:
+  # Default coding agent to launch from dashboard backlog (auto | claude | codex | agy | opencode | gemini).
+  # auto picks Claude Code if installed, otherwise the first agent found on PATH.
+  default: auto
+  # Per-tool model and effort for the session the board launches, passed as that
+  # tool's own flags (claude --model/--effort, agy --model/--effort, codex -m and
+  # model_reasoning_effort, gemini -m, opencode -m). Model names are the tool's own:
+  # `agy models`, `opencode models`, Claude's opus|sonnet|haiku. Absent = tool default.
+  # Uncomment the tool you launch:
+  # claude: { model: opus, effort: high }
+  # codex: { model: gpt-5, effort: low }
+  # agy: { model: gemini-3.1-pro-low }
+  # opencode: { model: anthropic/claude-sonnet-4-6 }
+  # gemini: { model: gemini-2.5-pro }
+