Skills › Compound Engineering › ce-setup › config-template.yaml

config-template.yaml

210 lines · skills/ce-setup/references/config-template.yaml at 3449230

Open on GitHub

1# Compound Engineering -- repo config

2# Copy to .compound-engineering/config.yaml in your project root (team defaults).

3# Optional override: .compound-engineering/config.local.yaml, same keys.

4# Ordinary keys: local wins when set, then this file, then the skill default.

5# docs_root is repo layout -- set it only here; local docs_root is ignored.

6# All settings are optional. Invalid ordinary values continue to the next layer.

7# docs_root fails closed on an unusable value.

9# --- Artifact root ---

10# Relocate every CE-written artifact folder (solutions, plans, ideation,

11# explainers, pulse-reports, dogfood-reports,

12# feedback-sweep, personas) under one repo-relative root. Unset -> `docs`,

13# byte-identical to today. A configured root becomes the sole location CE reads

14# and writes; CE never falls back to reading `docs/`.

15#

16# Set this only in this tracked file so every clone and worktree share one

17# artifact tree. A docs_root in config.local.yaml is ignored.

18#

19# The value must be a repo-relative directory whose real (symlink-resolved) path

20# stays inside the repo, is not the repo root itself, and is not under `.git/`.

21# Unlike every other setting here, an unusable value does NOT fall through to the

22# default -- CE stops with an error so a typo can't silently write artifacts into

23# the location you configured away from. `/ce-setup` reports the resolved root.

25# docs_root: .compound-engineering/artifacts # repo-relative dir (default: docs)

27# --- Product pulse ---

28# Settings written by /ce-product-pulse first-run interview. Re-run the skill with

29# argument `setup` or `reconfigure` to edit interactively.

31# pulse_product_name: "Spiral" # used in report titles (no default)

32# pulse_lookback_default: 24h # 1h | 24h | 7d | 30d (default: 24h)

33# pulse_primary_event: "session_started" # the event that means "user showed up"

34# pulse_value_event: "task_completed" # the event that means "user got value"

35# pulse_completion_events: "onboarded,first_purchase" # comma-separated, 0-3 events

36# pulse_quality_scoring: false # true | false (default: false; AI products only)

37# pulse_quality_dimension: "answer accuracy" # dimension scored 1-5 when pulse_quality_scoring is true

38# pulse_analytics_source: posthog # posthog | mixpanel | custom (no default)

39# pulse_tracing_source: sentry # sentry | datadog | custom (no default)

40# pulse_payments_source: stripe # stripe | custom (no default)

41# pulse_db_enabled: false # true | false (default: false; read-only DB if true)

42# pulse_metric_sources: "retention_d7=posthog,nps=delighted" # strategy-metric -> source overrides; comma-separated 'metric=source' pairs; unlisted metrics fall back to pulse_analytics_source

43# pulse_pending_metrics: "retention_d7,nps" # comma-separated strategy metrics awaiting instrumentation; render as 'no data'

44# pulse_excluded_metrics: "north_star" # comma-separated strategy metrics intentionally not in pulse

46# --- Output format ---

47# Per-skill output format default. Selects the exclusive format the artifact

48# is written in: `md` produces a markdown file, `html` produces a single

49# self-contained HTML file. The two are mutually exclusive -- there is no

50# sibling artifact. See DESIGN.md or your agent instructions to influence

51# HTML styling. Precedence: a format request in your prompt for that run wins

52# (e.g. "output:html" or "make it HTML"); a preference you established earlier

53# (in-session, saved to memory, or in your agent instructions) overrides these

54# keys (this config is the persisted fallback). A headless or pipeline run

55# resolves the format the same way; nothing forces `md`.

57# plan_output: html # md | html (default: md)

58# brainstorm_output: html # md | html (default: md)

59# ideate_output: md # md | html (default: html -- ideation docs are human-facing, so HTML is the default; set md to opt out)

61# --- Model elevation ---

62# Send only the reasoning-heavy step of /ce-plan (interpret research, then author

63# the plan) and /ce-brainstorm (generate approaches) to a model you name. The rest

64# of each skill -- dialogue, research, orchestration -- stays on your session

65# model. The value is a model alias (e.g. `fable`, `opus`). You can also name a

66# model in a prompt for a one-off, which overrides these keys.

67#

68# Cross-harness: setting a key means every harness you run these skills in will

69# attempt that model. The host serves it natively where it can; otherwise the

70# Claude CLI is invoked when installed and authenticated; otherwise the step runs

71# on your session model and the run says so. Invalid/commented/missing -> no

72# elevation.

74# plan_model: fable # model alias for plan authoring; candidate preference when a Bake-off runs (no default)

75# brainstorm_model: fable # model alias for approach generation; candidate preference when a Bake-off runs (no default)

77# --- PR concept teaching (ce-commit-push-pr) ---

78# When a PR introduces a concept new to the codebase, the generated PR

79# description gains a "New concepts" section teaching it. The section toggle is

80# the single gate: off disables judgment, the section, the /ce-explain offer,

81# and archival. Archival additionally writes the explainer to docs/explainers/

82# and links it from the PR (full-workflow runs only; override per run with

83# archive:on|off). Ordinary cascade: local, then this file, then the skill default.

85# pr_teaching_section: false # true | false (default: true)

86# pr_teaching_archive: true # true | false (default: false; full-workflow runs only)

88# --- ce-commit-push-pr babysit handoff ---

89# After a full-workflow PR is opened, ce-commit-push-pr auto-invokes ce-babysit-pr

90# to watch CI + incoming review and drive the PR toward merge-ready. Set false to

91# opt out standing (per-run override: the babysit:off token). This governs that

92# standing watch only -- lfg's in-pipeline babysit is bounded (3 fix rounds) and

93# still runs, since it is how that pipeline reaches "CI decided".

94# auto_babysit: false # true | false (default: true)

96# --- ce-plan scoping confirmation ---

97# By default /ce-plan pauses before research/plan-write to show a scoping

98# summary and wait for you to confirm scope. Set this to skip that wait: ce-plan

99# composes the summary for itself, records any inferred scope under an

100# Assumptions section, announces that it's proceeding, and keeps going. It skips

101# only that confirmation -- genuine blocking questions and the post-plan menu

102# still appear. Override per run with `confirm:auto` (skip) or `confirm:ask`

103# (force the gate on for one run).

104

105# plan_skip_scoping_confirm: true # true | false (default: false)

106

107# --- Cross-model review policy and preference ---

108# `cross_model_review_mode` decides whether ce-doc-review and ce-code-review may

109# automatically send review content (diffs, documents) to a second model

110# provider when their conditional cross-model pass activates. `off` skips that

111# pass before any route is chosen or process starts, leaves every local

112# reviewer in place, and reports "disabled by checkout config" instead of an

113# availability failure. `auto` keeps the current behavior. A direct request in

114# conversation ("use codex as the independent reviewer this time") overrides

115# `off` for that run; a conversation prohibition overrides `auto`.

116#

117# `cross_model_peer` is the preferred peer, consulted only when the mode is

118# `auto`. Conversation instructions still win, and the skills skip routes that

119# cannot establish the required host/peer distinction.

120# `cross_model_model` and `cross_model_effort` pin the model and reasoning

121# effort of whichever peer target runs; unset keeps the skills' editorial

122# mapping. The model must belong to that target's family (an alias such as

123# `fable` or a full id such as `claude-opus-5-5` for claude; `gpt-*` for codex,

124# optionally namespace-qualified such as `openai.gpt-6.1-sol` when the codex CLI

125# routes through a non-default `model_provider`; a `gemini-*` id such as

126# `gemini-3.8-flash` for antigravity).

128# xhigh plus max and ultra on the models that list them, grok low|medium|high|

130# max as each model lists them; cursor/composer routes take none. A value the peer

131# cannot honor skips the pass with a stated reason -- never substituted.

132

133# cross_model_review_mode: off # auto | off (default: auto)

135# cross_model_model: fable # model alias or full id for the resolved peer target (no default)

136# cross_model_effort: xhigh # reasoning effort for the resolved peer target (no default)

137

138# --- ce-work implementation engine ---

139# Standing preferences for who authors implementation units. Live

140# task/session/project instructions and LFG's provenance-bearing binding may

141# override or narrow these defaults. `off` disables only this standing preference.

142# `prefer` tries candidates in order, then falls back natively with disclosure.

143# `require` keeps the requested external identity fixed while viable, then also

144# falls back natively with disclosure if the route is unavailable. Harnesses:

146# default. Composer is a model

147# family through Cursor: { harness: cursor, model: composer }. Do not put CLI flags

148# or commands here. A candidate equivalent to the current host/default is skipped;

149# a different explicit model in the same harness remains eligible.

150

151# work_engine_mode: prefer # off | prefer | require (default: off)

152# work_engine_preferences:

153# - harness: cursor

154# model: composer

155# - harness: codex

156# model: "gpt-6.1-sol"

157# - harness: claude # configured default model

158

159# Reasoning effort for the external worker, per harness. Each value is one of

160# that harness's own effort levels. An unset harness keeps its own default:

161# Codex and Claude at high, native Grok at xhigh, Antigravity at high (it takes

162# low|medium|high). OpenCode (2.x) keeps its own per-model default and takes

164# effort. A map in config.local.yaml replaces this one.

165

166# work_engine_effort:

167# codex: xhigh

168# claude: max

169

170# --- ce-promote ---

171# Written automatically when you decline the Spiral setup offer in /ce-promote.

172# Suppresses that one-time setup nudge in this project. Remove the key to re-enable.

173

174# ce_promote_spiral_optout: true # true | (absent) (default: absent -- offer once)

175

176# --- Sweep (ce-sweep) ---

177

178# Feedback sources swept by /ce-sweep. Generic key: other skills may read this list.

179# Each entry: type (slack | github-issues | email), id (short name), target (channel ID /

180# owner/repo / mailbox), ack_action, closeout_action, sensitive (true withholds content

181# from committed state/plan text), approved (standing approval for source-side writes).

182# feedback_sources:

183# - { type: slack, id: slack-alpha, target: C0XXXXXXX, ack_action: eyes, closeout_action: white_check_mark, sensitive: false, approved: true }

184# - { type: github-issues, id: gh-issues, target: owner/repo, ack_action: "feedback:ack", closeout_action: "feedback:resolved", sensitive: false, approved: true }

185

186# Sweep-owned settings.

187# sweep_state_path: docs/feedback-sweep/state.yml # committed (multi-agent) or /tmp path (solo)

188# sweep_ack_cap: 25 # max acks per source per run before the circuit breaker

189# sweep_lease_ttl_minutes: 60 # single-writer lease staleness threshold

190# sweep_shared_branch: false # true: push-gated lease for shared-docs-branch topology

191

192# --- Compound Packs ---

193

194# Prescriptive domain knowledge folders that planning reads and cites.

195# Declared, never scanned: an entry names a source (repo-relative path,

196# ~/absolute path, or git URL) and what to install from it. Lists from this

197# file and config.local.yaml concatenate -- local adds packs, never replaces

198# the team's. Git sources require ref (tag or sha reproduce exactly; a branch

199# freezes at its cached resolution and can drift across machines). path: and

200# pasted GitHub /tree/<ref>/<sub> URLs scope a source to a subfolder.

201# pack: picks one id or a list; omit it to install everything the source

202# publishes. id: renames a single-pack entry.

203# Rules are top-level .md files with title + applies_when; subfolders are storage.

204

205# packs:

206# - source: compound-packs/local-rules # repo-relative, read live

207# - source: ~/compound-packs/kk-style # machine-local, read live

208# - source: https://github.com/org/rails-ce-pack # git, cached at ref

209# ref: v1.2.0

210# pack: [rails, inertia]