repo-harness 0.21.2 → 0.22.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/AGENTS.md +1 -0
- package/CLAUDE.md +1 -0
- package/README.es.md +39 -66
- package/README.fr.md +41 -67
- package/README.ja.md +39 -64
- package/README.md +40 -67
- package/README.zh-CN.md +34 -61
- package/SKILL.md +4 -0
- package/agents/fleet/deep-worker.md +6 -5
- package/agents/fleet/fast-worker.md +6 -5
- package/assets/hooks/lib/workflow-state.sh +1 -1
- package/assets/reference-configs/agentic-development-flow.md +2 -0
- package/assets/reference-configs/document-generation.md +2 -3
- package/assets/reference-configs/external-tooling.md +83 -158
- package/assets/reference-configs/global-working-rules.md +8 -0
- package/assets/reference-configs/harness-overview.md +2 -5
- package/assets/reference-configs/heartbeat-triage.md +1 -3
- package/assets/reference-configs/herdr-notify.md +6 -0
- package/assets/reference-configs/hook-operations.md +22 -10
- package/assets/reference-configs/host-invariants.md +1 -1
- package/assets/skill-commands/manifest.json +26 -3
- package/assets/skill-commands/repo-harness-architecture/SKILL.md +3 -3
- package/assets/skill-commands/repo-harness-architecture/references/worker.md +59 -32
- package/assets/skill-commands/repo-harness-check/SKILL.md +3 -0
- package/assets/skill-commands/repo-harness-check/references/architecture-evidence.md +70 -0
- package/assets/skill-version.json +2 -2
- package/assets/skills/repo-harness-cross-review/SKILL.md +4 -2
- package/assets/skills/repo-harness-cross-review/references/generic-review.md +30 -0
- package/assets/skills/windows-python-first/SKILL.md +76 -0
- package/assets/skills/windows-python-first/agents/openai.yaml +4 -0
- package/assets/templates/helpers/archive-workflow.sh +0 -6
- package/assets/templates/helpers/capability-resolver.ts +148 -110
- package/assets/templates/helpers/contract-worktree.sh +5 -5
- package/assets/templates/helpers/heartbeat-triage.sh +0 -18
- package/assets/templates/helpers/merge-gate.ts +12 -2
- package/assets/templates/helpers/refresh-current-status.sh +1 -1
- package/assets/templates/helpers/workflow-contract.ts +1 -1
- package/assets/templates/helpers/workstream-sync.sh +7 -4
- package/assets/workflow-contract.v1.json +33 -12
- package/dist/hook-entry.js +8283 -10503
- package/dist/oar-coding-host.js +13 -2
- package/dist/oar-review-host.js +14 -0
- package/dist/operator-ui/assets/ArchitectureWorkspace-D4OUgv7P.js +1 -0
- package/dist/operator-ui/assets/index-DAePh_VG.css +2 -0
- package/dist/operator-ui/assets/index-y71_oLs9.js +18 -0
- package/dist/operator-ui/index.html +6 -6
- package/package.json +11 -9
- package/references/bot-control.md +53 -0
- package/scripts/archive-workflow.sh +0 -6
- package/scripts/axr5-archctx-clean-room.ts +3 -3
- package/scripts/capability-resolver.ts +51 -30
- package/scripts/check-ci.sh +21 -1
- package/scripts/check-pi-package-smoke.ts +1 -1
- package/scripts/check-state-boundaries.ts +3 -2
- package/scripts/check-tarball-install-smoke.sh +21 -9
- package/scripts/contract-worktree.sh +5 -5
- package/scripts/create-project-dirs.sh +0 -1
- package/scripts/heartbeat-triage.sh +0 -18
- package/scripts/init-project.sh +0 -1
- package/scripts/lib/project-init-lib.sh +6 -49
- package/scripts/me2b-runtime-admission-canary.ts +4 -4
- package/scripts/merge-gate.ts +12 -2
- package/scripts/refresh-current-status.sh +1 -1
- package/scripts/run-harness-profile-benchmark.ts +1 -1
- package/scripts/test-file-durations.json +0 -41
- package/scripts/workflow-contract.ts +1 -1
- package/scripts/workstream-sync.sh +7 -4
- package/src/cli/commands/architecture-configuration.ts +20 -5
- package/src/cli/commands/architecture-projection.ts +17 -127
- package/src/cli/commands/capability-context.ts +17 -186
- package/src/cli/commands/doctor.ts +57 -20
- package/src/cli/commands/global-runtime.ts +38 -7
- package/src/cli/commands/init-hook.ts +70 -28
- package/src/cli/commands/init.ts +1 -4
- package/src/cli/commands/operator.ts +5 -2
- package/src/cli/commands/pm.ts +2 -3
- package/src/cli/commands/refactor.ts +61 -196
- package/src/cli/commands/review.ts +6 -3
- package/src/cli/commands/run.ts +0 -4
- package/src/cli/commands/setup-check-child.ts +8 -0
- package/src/cli/commands/setup-snapshot.ts +70 -0
- package/src/cli/commands/task-agent.ts +7 -0
- package/src/cli/hook/handler-registry.ts +0 -26
- package/src/cli/hook/mutation-observed.ts +5 -123
- package/src/cli/hook/session-context-budget.ts +0 -3
- package/src/cli/hook/session-context.ts +28 -164
- package/src/cli/hook/stop-handler.ts +0 -38
- package/src/cli/hook/subagent-handler.ts +3 -1
- package/src/cli/hook-entry.ts +0 -4
- package/src/cli/index.ts +2 -7
- package/src/cli/mcp/pm-server.ts +9 -6
- package/src/cli/mcp/transports/http.ts +9 -0
- package/src/core/adoption/standard-plan.ts +31 -16
- package/src/core/architecture/projection.ts +34 -551
- package/src/core/capabilities/registry.ts +98 -79
- package/src/core/dev-activity/decode.ts +204 -0
- package/src/core/dev-activity/projection.ts +349 -0
- package/src/core/dev-activity/types.ts +147 -0
- package/src/core/hook-work-budget.ts +0 -2
- package/src/core/pm/protocol.ts +15 -0
- package/src/core/refactor/provider-contract.ts +15 -180
- package/src/core/setup/decode.ts +174 -0
- package/src/core/setup/projection.ts +212 -0
- package/src/core/setup/types.ts +154 -0
- package/src/effects/architecture/archctx-provider.ts +22 -283
- package/src/effects/architecture/model-reader.ts +9 -15
- package/src/effects/architecture/projection-apply.ts +79 -0
- package/src/effects/architecture/projection-config.ts +19 -6
- package/src/effects/capabilities/archcontext-owners.ts +95 -0
- package/src/effects/dev-activity/collector.ts +507 -0
- package/src/effects/engineers/profile-store.ts +2 -11
- package/src/effects/locking/exclusive-directory-lock.ts +3 -0
- package/src/effects/operator/fleet-collector-process.ts +6 -15
- package/src/effects/operator/server.ts +73 -47
- package/src/effects/pipeline/authority.ts +1 -0
- package/src/effects/pipeline/store.ts +27 -3
- package/src/effects/pm/operations.ts +4 -3
- package/src/effects/publication/merge-readiness.ts +6 -3
- package/src/effects/refactor/archctx-provider.ts +32 -82
- package/src/effects/refactor/recommendations.ts +87 -99
- package/src/effects/review/generic-review.ts +102 -25
- package/src/effects/review/oar-review-host.ts +13 -3
- package/src/effects/runtime/helper-runner.ts +2 -16
- package/src/effects/runtime/protected-helper-platform.ts +34 -1
- package/src/effects/setup/collector.ts +164 -0
- package/src/effects/state/resolve-effective-state.ts +38 -11
- package/src/effects/terminal/task-request-readback.ts +79 -0
- package/src/effects/terminal/task-session.ts +125 -27
- package/src/operator-web/App.tsx +391 -1674
- package/src/operator-web/ArchitectureWorkspace.tsx +2 -18
- package/src/operator-web/Board.tsx +436 -0
- package/src/operator-web/Drawer.tsx +113 -0
- package/src/operator-web/ItemDrawer.tsx +183 -0
- package/src/operator-web/OrganizationSummary.tsx +1 -1
- package/src/operator-web/PipelineBoard.tsx +11 -7
- package/src/operator-web/PlanningView.tsx +1 -1
- package/src/operator-web/RepositoriesPage.tsx +90 -0
- package/src/operator-web/RepositoryWorkspace.tsx +1097 -0
- package/src/operator-web/RuntimeBadges.tsx +1 -1
- package/src/operator-web/SetupPages.tsx +444 -0
- package/src/operator-web/SystemPage.tsx +99 -0
- package/src/operator-web/app.css +207 -0
- package/src/operator-web/clipboard.ts +13 -0
- package/src/operator-web/css.d.ts +2 -0
- package/src/operator-web/dev-activity.ts +155 -0
- package/src/operator-web/fixture.ts +0 -38
- package/src/operator-web/fixtures/dev-activity.ts +195 -0
- package/src/operator-web/fixtures/index.ts +99 -0
- package/src/operator-web/fixtures/setup.ts +176 -0
- package/src/operator-web/fleet-api.ts +132 -0
- package/src/operator-web/i18n.ts +538 -130
- package/src/operator-web/index.html +4 -4
- package/src/operator-web/labels.ts +34 -0
- package/src/operator-web/main.tsx +34 -6
- package/src/operator-web/marks.tsx +4 -88
- package/src/operator-web/setup.ts +120 -0
- package/src/operator-web/theme.ts +71 -0
- package/src/operator-web/ui.tsx +94 -0
- package/src/operator-web/worklist.ts +131 -0
- package/src/operator-web/workspace-location.ts +43 -12
- package/src/pi/extension.ts +5 -2
- package/src/pi/hook-bridge.ts +33 -13
- package/assets/templates/helpers/architecture-event.ts +0 -1411
- package/assets/templates/helpers/architecture-queue.sh +0 -709
- package/assets/templates/helpers/archive-architecture-request.sh +0 -423
- package/assets/templates/helpers/check-architecture-sync.sh +0 -436
- package/dist/operator-ui/assets/ArchitectureWorkspace-CBlfSPBR.js +0 -1
- package/dist/operator-ui/assets/ibm-plex-sans-latin-400-normal-CDDApCn2.woff2 +0 -0
- package/dist/operator-ui/assets/ibm-plex-sans-latin-400-normal-CYLoc0-x.woff +0 -0
- package/dist/operator-ui/assets/ibm-plex-sans-latin-500-normal-6ng42L7E.woff2 +0 -0
- package/dist/operator-ui/assets/ibm-plex-sans-latin-500-normal-BgVn5rGT.woff +0 -0
- package/dist/operator-ui/assets/ibm-plex-sans-latin-600-normal-Cu4Hd6ag.woff +0 -0
- package/dist/operator-ui/assets/ibm-plex-sans-latin-600-normal-CuJfVYMP.woff2 +0 -0
- package/dist/operator-ui/assets/ibm-plex-sans-latin-700-normal-Bth3BMcD.woff +0 -0
- package/dist/operator-ui/assets/ibm-plex-sans-latin-700-normal-Bxkt5Cjx.woff2 +0 -0
- package/dist/operator-ui/assets/index-BCD_MgSs.js +0 -9
- package/dist/operator-ui/assets/index-CRN670y2.css +0 -1
- package/dist/operator-ui/assets/jetbrains-mono-latin-400-normal-6-qcROiO.woff +0 -0
- package/dist/operator-ui/assets/jetbrains-mono-latin-400-normal-V6pRDFza.woff2 +0 -0
- package/dist/operator-ui/assets/jetbrains-mono-latin-500-normal-BWZEU5yA.woff2 +0 -0
- package/dist/operator-ui/assets/jetbrains-mono-latin-500-normal-CJOVTJB7.woff +0 -0
- package/dist/operator-ui/assets/jetbrains-mono-latin-700-normal-BYuf6tUa.woff2 +0 -0
- package/dist/operator-ui/assets/jetbrains-mono-latin-700-normal-D3wTyLJW.woff +0 -0
- package/dist/operator-ui/assets/space-grotesk-latin-400-normal-BnQMeOim.woff +0 -0
- package/dist/operator-ui/assets/space-grotesk-latin-400-normal-CJ-V5oYT.woff2 +0 -0
- package/dist/operator-ui/assets/space-grotesk-latin-500-normal-CNSSEhBt.woff +0 -0
- package/dist/operator-ui/assets/space-grotesk-latin-500-normal-lFbtlQH6.woff2 +0 -0
- package/dist/operator-ui/assets/space-grotesk-latin-600-normal-BflQw4A9.woff +0 -0
- package/dist/operator-ui/assets/space-grotesk-latin-600-normal-DjKNqYRj.woff2 +0 -0
- package/dist/operator-ui/assets/space-grotesk-latin-700-normal-CwsQ-cCU.woff +0 -0
- package/dist/operator-ui/assets/space-grotesk-latin-700-normal-RjhwGPKo.woff2 +0 -0
- package/scripts/architecture-event.ts +0 -1411
- package/scripts/architecture-queue.sh +0 -709
- package/scripts/archive-architecture-request.sh +0 -423
- package/scripts/check-architecture-sync.sh +0 -436
- package/src/cli/commands/automation.ts +0 -250
- package/src/cli/commands/refactor-recommendation-configuration.ts +0 -13
- package/src/cli/hook/architecture-drift.ts +0 -345
- package/src/core/architecture/restamp-publication.ts +0 -94
- package/src/core/automation/budget.ts +0 -1584
- package/src/core/automation/controller.ts +0 -246
- package/src/core/automation/projection.ts +0 -201
- package/src/core/operator/automation-summary.ts +0 -72
- package/src/core/operator/repository-snapshot.ts +0 -39
- package/src/core/refactor/activation.ts +0 -42
- package/src/core/refactor/architecture-intervention.ts +0 -49
- package/src/core/refactor/board.ts +0 -54
- package/src/core/refactor/candidate-verification.ts +0 -49
- package/src/core/refactor/execution-binding.ts +0 -61
- package/src/core/refactor/materialization.ts +0 -154
- package/src/core/refactor/policy.ts +0 -125
- package/src/core/refactor/program-state.ts +0 -209
- package/src/core/refactor/program.ts +0 -178
- package/src/core/refactor/proposal-authoring.ts +0 -78
- package/src/core/refactor/workflow-route.ts +0 -76
- package/src/effects/architecture/projection-acceptance.ts +0 -879
- package/src/effects/architecture/projection-continuation.ts +0 -55
- package/src/effects/architecture/projection-jobs.ts +0 -579
- package/src/effects/architecture/projection-orchestrator.ts +0 -280
- package/src/effects/architecture/refresh-consumer.ts +0 -119
- package/src/effects/architecture/restamp-publication.ts +0 -293
- package/src/effects/automation/budget-store.internal.ts +0 -15
- package/src/effects/automation/budget-store.ts +0 -1861
- package/src/effects/automation/clock.ts +0 -91
- package/src/effects/automation/controller-run.ts +0 -264
- package/src/effects/automation/controller-store.ts +0 -165
- package/src/effects/automation/grant-store.ts +0 -226
- package/src/effects/operator/automation-summary.ts +0 -95
- package/src/effects/refactor/activation-store.ts +0 -17
- package/src/effects/refactor/architecture-intervention.ts +0 -54
- package/src/effects/refactor/candidate-verification.ts +0 -127
- package/src/effects/refactor/discovery-authoring.ts +0 -109
- package/src/effects/refactor/execution-binding-store.ts +0 -54
- package/src/effects/refactor/materialization.ts +0 -158
- package/src/effects/refactor/post-merge-resolution.ts +0 -155
- package/src/effects/refactor/program-store.ts +0 -198
- package/src/effects/refactor/recommendation-settings.ts +0 -13
- package/src/effects/refactor/resolution-store.ts +0 -24
- package/src/effects/refactor/shadow-discovery.ts +0 -132
- package/src/operator-web/AutomationSummary.tsx +0 -134
- package/src/operator-web/icons.tsx +0 -80
- package/src/operator-web/repository-snapshot.ts +0 -27
- package/src/operator-web/styles.css +0 -523
- package/src/operator-web/styles.d.ts +0 -4
package/AGENTS.md
CHANGED
|
@@ -13,6 +13,7 @@
|
|
|
13
13
|
- `start` recovers only a real target directory registered by this clone on the expected branch, under the topology lock, with no lock or other owner. Recovery requires a current-user-owned checkout directory without group or other write access, a regular owned Git pointer, its matching admin backpointer, and the original directory identity saved by start. Git metadata can be group-writable but must not be other-writable. Start still creates a checkout when uid is unavailable; recovery then refuses. Exact removal retires the identity after Git readback. Missing identity or a replacement directory blocks recovery. Files, symlinks, unregistered paths and other branches are refused. Cleanup refuses a registered worktree whose branch is missing. MCP cleanup also refuses active markers and retained native verification evidence. Sweep reports name processed paths. Deferred entries are reported as counts. `kept` excludes deferred entries. A command timeout sends SIGTERM to the direct child only. A grandchild can briefly extend the deadline. No detached cleanup process starts. Herdr notifications accept registered default task worktrees and paths inside them. Other temporary paths and `rh-herdr-` labels stay filtered.
|
|
14
14
|
- Automatically delete only merged, clean worktrees/branches; ask the user for every other deletion. Credentials/permissions (including confirmation bypass) and release/production operations require user approval.
|
|
15
15
|
- Use gatekeeper/cross-model review for large changes, security/permissions, or unresolved model uncertainty; ordinary steps record diagnostics without approval loops.
|
|
16
|
+
- When you delegate, investigate first, then split work by write ownership. Give each parallel writer its own worktree from a clean source. Implementation sub-agents edit only their scope. They do not run build, typecheck or tests. They commit early in their own worktree and return early. In a shared checkout they do not commit; the parent commits. Evidence roles run the named check that is their task and do not commit. The parent agent owns task management, conflict resolution, build, typecheck, tests and integration. After integration, the parent removes each sub-agent worktree and branch whose commits are in the task branch and whose checkout is clean. It also removes the intermediate build output and temporary files that the work created. Other deletion still needs the user.
|
|
16
17
|
- Model division and cross-model dispatch/review follow [Herdr Dispatch](docs/reference-configs/external-tooling.md#herdr-dispatch).
|
|
17
18
|
- GitHub operations (PRs, reviews, merges, labels, releases, comments) go only through the gh/git CLI or the GitHub API. Never use simulated browser clicks on github.com: the account can be banned.
|
|
18
19
|
- When one PR changes both test assertions and implementation code, run one read-only review focused on tests bent to fit a bug (on-demand, not a restored gate; dispatch per the herdr guide).
|
package/CLAUDE.md
CHANGED
|
@@ -13,6 +13,7 @@
|
|
|
13
13
|
- `start` recovers only a real target directory registered by this clone on the expected branch, under the topology lock, with no lock or other owner. Recovery requires a current-user-owned checkout directory without group or other write access, a regular owned Git pointer, its matching admin backpointer, and the original directory identity saved by start. Git metadata can be group-writable but must not be other-writable. Start still creates a checkout when uid is unavailable; recovery then refuses. Exact removal retires the identity after Git readback. Missing identity or a replacement directory blocks recovery. Files, symlinks, unregistered paths and other branches are refused. Cleanup refuses a registered worktree whose branch is missing. MCP cleanup also refuses active markers and retained native verification evidence. Sweep reports name processed paths. Deferred entries are reported as counts. `kept` excludes deferred entries. A command timeout sends SIGTERM to the direct child only. A grandchild can briefly extend the deadline. No detached cleanup process starts. Herdr notifications accept registered default task worktrees and paths inside them. Other temporary paths and `rh-herdr-` labels stay filtered.
|
|
14
14
|
- Automatically delete only merged, clean worktrees/branches; ask the user for every other deletion. Credentials/permissions (including confirmation bypass) and release/production operations require user approval.
|
|
15
15
|
- Use gatekeeper/cross-model review for large changes, security/permissions, or unresolved model uncertainty; ordinary steps record diagnostics without approval loops.
|
|
16
|
+
- When you delegate, investigate first, then split work by write ownership. Give each parallel writer its own worktree from a clean source. Implementation sub-agents edit only their scope. They do not run build, typecheck or tests. They commit early in their own worktree and return early. In a shared checkout they do not commit; the parent commits. Evidence roles run the named check that is their task and do not commit. The parent agent owns task management, conflict resolution, build, typecheck, tests and integration. After integration, the parent removes each sub-agent worktree and branch whose commits are in the task branch and whose checkout is clean. It also removes the intermediate build output and temporary files that the work created. Other deletion still needs the user.
|
|
16
17
|
- Model division and cross-model dispatch/review follow [Herdr Dispatch](docs/reference-configs/external-tooling.md#herdr-dispatch).
|
|
17
18
|
- GitHub operations (PRs, reviews, merges, labels, releases, comments) go only through the gh/git CLI or the GitHub API. Never use simulated browser clicks on github.com: the account can be banned.
|
|
18
19
|
- When one PR changes both test assertions and implementation code, run one read-only review focused on tests bent to fit a bug (on-demand, not a restored gate; dispatch per the herdr guide).
|
package/README.es.md
CHANGED
|
@@ -22,10 +22,10 @@ proyecto, de modo que la siguiente sesión de agente continúa desde archivos en
|
|
|
22
22
|
lugar del historial de chat. Adopta un repositorio existente con un contract
|
|
23
23
|
de agente tasks-first que mantiene alineados a Claude y Codex.
|
|
24
24
|
|
|
25
|
-
Sobre ese contract
|
|
26
|
-
|
|
27
|
-
|
|
28
|
-
|
|
25
|
+
Sobre ese contract, el Bot planifica **programas de larga duración**: los task
|
|
26
|
+
offers y los leases permiten que un Sprint avance a través de sesiones.
|
|
27
|
+
repo-harness aporta las herramientas y la evidencia; el Bot decide qué se
|
|
28
|
+
ejecuta a continuación.
|
|
29
29
|
|
|
30
30
|
## Índice
|
|
31
31
|
|
|
@@ -172,10 +172,9 @@ repo-harness mcp uninstall --services-stopped # after stopping all MCP HTTP serv
|
|
|
172
172
|
decisión humana cabe en una sola pantalla — verdict, archivos previstos vs
|
|
173
173
|
reales, comandos que pasaron, riesgo residual, rollback — en lugar de una
|
|
174
174
|
reconstrucción de lo que el agente afirma haber hecho.
|
|
175
|
-
- **El trabajo
|
|
176
|
-
|
|
177
|
-
|
|
178
|
-
aceptación sin un receipt. La autonomía está acotada por artefactos, no por
|
|
175
|
+
- **El trabajo planificado sigue siendo responsable.** Una tarea no puede
|
|
176
|
+
retenerse más allá de su lease. Tampoco puede reclamar aceptación sin un
|
|
177
|
+
receipt. El Bot planifica a partir de evidencia registrada, no de la
|
|
179
178
|
confianza.
|
|
180
179
|
|
|
181
180
|
En un repositorio adoptado, la superficie se mantiene intencionalmente
|
|
@@ -199,21 +198,20 @@ completo para un repositorio en solitario, y todo lo que está en
|
|
|
199
198
|
[Flujo de trabajo de tareas](#flujo-de-trabajo-de-tareas) pertenece aquí. Nada
|
|
200
199
|
de lo que sigue es necesario para usarlo.
|
|
201
200
|
|
|
202
|
-
**Capa 2 — programas
|
|
203
|
-
una sesión: un
|
|
204
|
-
|
|
205
|
-
|
|
206
|
-
|
|
207
|
-
y retiene el trabajo mediante leases renovables. Ver
|
|
201
|
+
**Capa 2 — programas planificados por el Bot.** Trabajo de larga duración que
|
|
202
|
+
sobrevive a una sesión: un plano de colaboración donde varios Module Engineers
|
|
203
|
+
intercambian señales y handoffs, y task offers de engineer retenidos mediante
|
|
204
|
+
leases renovables. El Bot decide qué planificar. repo-harness aporta las
|
|
205
|
+
herramientas y la evidencia. Ver
|
|
208
206
|
[Programas autorizados](#programas-autorizados).
|
|
209
207
|
|
|
210
208
|
| | Capa 1 | Capa 2 |
|
|
211
209
|
| --- | --- | --- |
|
|
212
|
-
| Unidad de trabajo | Un task contract |
|
|
213
|
-
| Quién lo conduce | Un humano en una sesión |
|
|
214
|
-
| Autoridad | Plan, contract, review, checks | Lo anterior, más
|
|
215
|
-
| Punto de entrada | `repo-harness init` | `repo-harness
|
|
216
|
-
| Condición de parada | Closeout de la tarea |
|
|
210
|
+
| Unidad de trabajo | Un task contract | Una engineer task planificada |
|
|
211
|
+
| Quién lo conduce | Un humano en una sesión | El Bot, a través del CLI de repo-harness y Herdr |
|
|
212
|
+
| Autoridad | Plan, contract, review, checks | Lo anterior, más engineer binding y lease |
|
|
213
|
+
| Punto de entrada | `repo-harness init` | `repo-harness engineer acquire-next` |
|
|
214
|
+
| Condición de parada | Closeout de la tarea | Closeout de la tarea o lease perdido |
|
|
217
215
|
|
|
218
216
|
La capa 2 no reemplaza a la capa 1: cada paso de un programa sigue
|
|
219
217
|
proyectándose en los mismos artefactos de plan, contract y review que habría
|
|
@@ -226,8 +224,7 @@ escrito un humano.
|
|
|
226
224
|
| **Sesiones respaldadas por archivos** | Plans, contracts, checks y handoffs viven en el repositorio, de modo que una sesión nueva retoma desde artefactos en vez de un hilo de chat |
|
|
227
225
|
| **Typed hook runtime** | Ocho managed routes compartidas, más tres delegation routes exclusivas de Codex, cada una atada a exactamente un typed handler in-process, con guards fail-closed en el límite de edición |
|
|
228
226
|
| **Plan → Contract → Review** | Un solo ciclo de vida desde el plan aprobado hasta el contract proyectado, el worktree aislado, la evidencia estructurada y un closeout revisable |
|
|
229
|
-
| **Programas
|
|
230
|
-
| **Controller desatendido acotado** | Un bucle de dispatch de Engineer bajo topes duros de pasos, duración y reintentos, que reserva budget antes de cada intento |
|
|
227
|
+
| **Programas planificados por el Bot** | Programas de collaboration con task offers de engineer y leases renovables; el Bot planifica, repo-harness aporta herramientas y evidencia |
|
|
231
228
|
| **Carga de contexto progresiva** | Un root context estable de ~12KB más capability contracts de ~1KB que solo se cargan para los archivos que realmente se están tocando |
|
|
232
229
|
| **Integración con CodeGraph** | Consultas estructurales (callers, callees, definitions) respondidas desde un índice pre-construido en vez de pasadas repetidas de grep-and-read |
|
|
233
230
|
| **MCP planner sidecar** | ChatGPT lee el estado real del repositorio y escribe artefactos de PRD/Sprint/Goal; Codex los ejecuta, sin acceso de escritura al código fuente por defecto |
|
|
@@ -306,9 +303,9 @@ flowchart TD
|
|
|
306
303
|
PreHooks -->|blocked| ScopeFix["Fix plan, contract, worktree, or scope"]
|
|
307
304
|
ScopeFix --> Implement
|
|
308
305
|
PreHooks -->|allowed| Changes["Code, docs, tests, or config changes"]
|
|
309
|
-
Changes --> PostHooks["Post-edit and post-bash hooks<br/>trace,
|
|
310
|
-
PostHooks -->
|
|
311
|
-
|
|
306
|
+
Changes --> PostHooks["Post-edit and post-bash hooks<br/>trace, handoff, check evidence"]
|
|
307
|
+
PostHooks --> ArchUpdate["Explicit architecture update<br/>archctx ChangeSet when responsibilities change<br/>architecture-projection apply"]
|
|
308
|
+
ArchUpdate --> Verify["Run verification<br/>tests plus repo workflow checks"]
|
|
312
309
|
|
|
313
310
|
Verify --> Checks["Structured evidence<br/>.ai/harness/evidence/events/log.jsonl<br/>.ai/harness/runs/*.json"]
|
|
314
311
|
Checks --> CheckReview["Evaluator review<br/>Waza /check -> review file"]
|
|
@@ -341,40 +338,15 @@ y [`workflow-orchestration.md`](docs/reference-configs/workflow-orchestration.md
|
|
|
341
338
|
|
|
342
339
|
## Programas autorizados
|
|
343
340
|
|
|
344
|
-
Un programa es trabajo que sobrevive a una sesión.
|
|
345
|
-
|
|
341
|
+
Un programa es trabajo que sobrevive a una sesión. El Bot lo planifica a
|
|
342
|
+
través del CLI de repo-harness y Herdr. repo-harness no mantiene un scheduler,
|
|
343
|
+
un budget ledger ni un controller propios.
|
|
346
344
|
|
|
347
|
-
```bash
|
|
348
|
-
repo-harness automation grant mint # store one operator ProgramAuthorizationV2
|
|
349
|
-
repo-harness automation grant list # digests held for this repository
|
|
350
|
-
repo-harness automation budget show # the enforceable per-goal ledger
|
|
351
|
-
repo-harness automation budget repair # seal a stopped or expired run's exhaustion receipt
|
|
352
|
-
```
|
|
353
|
-
|
|
354
|
-
- **Authorization.** Un `ProgramAuthorizationV2` emitido por un operador vive
|
|
355
|
-
en el gate store del home del harness. No existe una ruta de arranque sin
|
|
356
|
-
autenticar, y un programa nunca deriva su propio actor — el autor de cada
|
|
357
|
-
registro se resuelve desde `--authorization-id`.
|
|
358
|
-
- **Budget.** Los agent turns, acquisitions y runner invocations reservan el presupuesto antes del trabajo. `budget repair` conserva los límites.
|
|
359
345
|
- **Lease.** El trabajo retenido lleva un lease renovable con un intervalo de
|
|
360
346
|
renovación, un TTL máximo y un conjunto cerrado de fuentes de evidencia. Un
|
|
361
347
|
estado de liveness no probado exige atención en lugar de reclamarse en
|
|
362
348
|
silencio.
|
|
363
349
|
|
|
364
|
-
### Controller desatendido
|
|
365
|
-
|
|
366
|
-
```bash
|
|
367
|
-
repo-harness automation controller start --maximum-steps 20 --maximum-duration-ms 300000
|
|
368
|
-
repo-harness automation controller step
|
|
369
|
-
repo-harness automation controller status
|
|
370
|
-
repo-harness automation controller stop
|
|
371
|
-
```
|
|
372
|
-
|
|
373
|
-
Un bucle de dispatch de Engineer bajo topes duros, con backoff determinista y
|
|
374
|
-
un ledger acotado de reintentos por intento. Cada intento reserva budget antes
|
|
375
|
-
de quedar registrado, y un resultado proyectado fuera del enum cerrado no puede
|
|
376
|
-
contarse como satisfecho.
|
|
377
|
-
|
|
378
350
|
### Scheduling de Engineer
|
|
379
351
|
|
|
380
352
|
```bash
|
|
@@ -393,20 +365,17 @@ sobre un backlog más antiguo.
|
|
|
393
365
|
|
|
394
366
|
Campaign execution moved to the existing Bot skills on 2026-10-04. Use [repo-harness](SKILL.md) to dispatch and collect work through Herdr/OAR. Use [repo-harness-product](assets/skills/repo-harness-product/SKILL.md) for planning and [repo-harness-check](assets/skill-commands/repo-harness-check/SKILL.md) for scope and verification. repo-harness has no campaign runtime.
|
|
395
367
|
|
|
396
|
-
### Refactor
|
|
368
|
+
### Refactor recommendations
|
|
397
369
|
|
|
398
370
|
```bash
|
|
399
|
-
repo-harness refactor
|
|
400
|
-
repo-harness refactor
|
|
401
|
-
repo-harness refactor verify-candidate
|
|
402
|
-
repo-harness refactor board
|
|
371
|
+
repo-harness refactor recommendations # measured suggestions with evidence
|
|
372
|
+
repo-harness refactor decide <id> accept|defer|reject --reason "<text>"
|
|
403
373
|
```
|
|
404
374
|
|
|
405
|
-
|
|
406
|
-
|
|
407
|
-
|
|
408
|
-
|
|
409
|
-
active.
|
|
375
|
+
ArchContext mide problemas estructurales como ciclos y violaciones de
|
|
376
|
+
dirección. repo-harness muestra cada sugerencia con sus métricas y las
|
|
377
|
+
estadísticas de sus módulos, y registra la decisión del usuario. Nunca ejecuta
|
|
378
|
+
un refactor; una sugerencia aceptada sigue el flujo normal de plan y pull request.
|
|
410
379
|
|
|
411
380
|
### Plano de colaboración
|
|
412
381
|
|
|
@@ -446,6 +415,8 @@ repo-harness review status --contract tasks/contracts/<task>.contract.md
|
|
|
446
415
|
repo-harness review close --contract tasks/contracts/<task>.contract.md
|
|
447
416
|
```
|
|
448
417
|
|
|
418
|
+
En la app de escritorio de Claude sin Herdr, añada `--launch app-terminal` y omita `--herdr-endpoint`. La primera llamada imprime el comando fijo del host. Claude lo ejecuta una vez en una pestaña Terminal de la app y repite la misma ronda. OAR, Seatbelt y el Receipt no cambian.
|
|
419
|
+
|
|
449
420
|
Un task-agent fleet `deep-reasoner` persiste en un linked checkout dedicado y la sesión Herdr indicada, durante un máximo de tres rondas. Por defecto se elige el harness opuesto al owner; `--harness` permite elegirlo explícitamente. Sólo un ejecutable ausente antes de iniciar permite fallback anunciado. El Result de archivo pasa la validación de dominio y el writer/verifier Receipt generic-review; close verifica el Receipt otra vez. History es observación. Claude usa una allowlist de archivos resultado y fingerprints; aislamiento OS de producción e ingestión completa de paquetes grandes siguen sin verificar.
|
|
450
421
|
|
|
451
422
|
## Hooks
|
|
@@ -583,11 +554,12 @@ host mientras el CLI y los hooks poseen la ejecución.
|
|
|
583
554
|
| `repo-harness-product` | Modos PRD, Sprint y Goal para el product planning de upper-layer |
|
|
584
555
|
| `repo-harness-check` | Definir el plan. Revisar el plan. Verificar los resultados registrados. |
|
|
585
556
|
| `repo-harness-ship` | Valida worktrees terminados, hace push de branches y abre PRs |
|
|
586
|
-
| `repo-harness-architecture` |
|
|
557
|
+
| `repo-harness-architecture` | Modelo de architecture, docs proyectados y diagramas sin un refresh completo del harness |
|
|
587
558
|
| `repo-harness-cross-review` | Generic acceptance through persistent fleet deep-reasoner task-agent + Herdr review; direct advisory runtime retired |
|
|
588
559
|
| `repo-harness-chatgpt` | Consultas de Oracle browser/GPT Pro, setup del MCP Connector y bridge handoff; solo setup explícito |
|
|
589
560
|
| `obsidian-memory` | Bot: leer o guardar memoria por solicitud explícita. |
|
|
590
561
|
| `repo-harness-test` | Worker: pruebas del código fuente y fixtures reales. |
|
|
562
|
+
| `windows-python-first` | Worker: tareas de archivos, texto y shell en Windows nativo con Python. |
|
|
591
563
|
| `merge-gate` (externo) | Gate final de exact-candidate; repo-harness no distribuye ningún Skill de merge-gate — ver [external tooling](docs/reference-configs/external-tooling.md) |
|
|
592
564
|
|
|
593
565
|
Los Bot skills definen el alcance, la delegación y las decisiones. Los Worker skills ejecutan los pasos.
|
|
@@ -595,7 +567,8 @@ Los Bot skills definen el alcance, la delegación y las decisiones. Los Worker s
|
|
|
595
567
|
Bot: `repo-harness`, `repo-harness-check`, `repo-harness-product`,
|
|
596
568
|
`repo-harness-ship`, `obsidian-memory`, `repo-harness-cross-review`,
|
|
597
569
|
`repo-harness-chatgpt`. Worker: `repo-harness-setup`,
|
|
598
|
-
`repo-harness-test`,
|
|
570
|
+
`repo-harness-test`, `repo-harness-architecture`, and
|
|
571
|
+
`windows-python-first`.
|
|
599
572
|
`obsidian-memory`, and `repo-harness-ship` requieren una invocación explícita.
|
|
600
573
|
Las tareas comunes usan un brief corto y la descripción del PR. PRD, Sprint y Goal sirven para el trabajo de producto solicitado.
|
|
601
574
|
|
|
@@ -676,8 +649,8 @@ repositorio adopte la misma política.
|
|
|
676
649
|
|
|
677
650
|
## Versión actual
|
|
678
651
|
|
|
679
|
-
- Paquete npm: `repo-harness@0.
|
|
680
|
-
- Sello de workflow generado: `repo-harness@0.
|
|
652
|
+
- Paquete npm: `repo-harness@0.22.1`
|
|
653
|
+
- Sello de workflow generado: `repo-harness@0.22.1+template@0.22.1`
|
|
681
654
|
- Repositorio de GitHub: `Ancienttwo/repo-harness`
|
|
682
655
|
- Notas de versión e historial: [`docs/CHANGELOG.md`](docs/CHANGELOG.md)
|
|
683
656
|
|
package/README.fr.md
CHANGED
|
@@ -22,10 +22,10 @@ de review, afin que la session d'agent suivante reprenne à partir des fichiers
|
|
|
22
22
|
plutôt que de la mémoire de chat. Il adopte un dépôt existant avec un contrat
|
|
23
23
|
d'agent tasks-first qui maintient Claude et Codex alignés.
|
|
24
24
|
|
|
25
|
-
Par-dessus ce contrat,
|
|
26
|
-
|
|
27
|
-
|
|
28
|
-
|
|
25
|
+
Par-dessus ce contrat, le Bot ordonnance des **programmes de longue durée** :
|
|
26
|
+
les task offers et les leases permettent à un Sprint d'avancer d'une session à
|
|
27
|
+
l'autre. repo-harness fournit les outils et les preuves ; le Bot décide de ce
|
|
28
|
+
qui s'exécute ensuite.
|
|
29
29
|
|
|
30
30
|
## Sommaire
|
|
31
31
|
|
|
@@ -171,11 +171,10 @@ repo-harness mcp uninstall --services-stopped # after stopping all MCP HTTP serv
|
|
|
171
171
|
décision humaine tient sur un seul écran — verdict, fichiers prévus vs
|
|
172
172
|
réels, commandes passées, risque résiduel, rollback — plutôt qu'une
|
|
173
173
|
reconstruction de ce que l'agent prétend avoir fait.
|
|
174
|
-
- **Le travail
|
|
175
|
-
|
|
176
|
-
|
|
177
|
-
|
|
178
|
-
artifacts, pas par la confiance.
|
|
174
|
+
- **Le travail ordonnancé reste redevable.** Une tâche ne peut pas être
|
|
175
|
+
retenue au-delà de son lease. Elle ne peut pas non plus revendiquer une
|
|
176
|
+
acceptance sans receipt. Le Bot ordonnance à partir des preuves enregistrées,
|
|
177
|
+
pas de la confiance.
|
|
179
178
|
|
|
180
179
|
Dans un dépôt adopté, la surface à comprendre reste volontairement réduite :
|
|
181
180
|
|
|
@@ -197,21 +196,20 @@ entier pour un dépôt en solo, et tout ce qui figure dans
|
|
|
197
196
|
[Workflow des tâches](#workflow-des-tâches) appartient à cette couche. Rien de
|
|
198
197
|
ce qui suit n'est requis pour l'utiliser.
|
|
199
198
|
|
|
200
|
-
**Couche 2 — les programmes
|
|
201
|
-
à une session : un
|
|
202
|
-
|
|
203
|
-
|
|
204
|
-
|
|
205
|
-
ledger par goal, et retient le travail via des leases renouvelables. Voir
|
|
199
|
+
**Couche 2 — les programmes ordonnancés par le Bot.** Du travail de longue
|
|
200
|
+
durée qui survit à une session : un plan de collaboration où plusieurs Module
|
|
201
|
+
Engineers échangent signaux et handoffs, et des task offers d'engineer retenus
|
|
202
|
+
via des leases renouvelables. Le Bot décide de ce qu'il ordonnance. repo-harness
|
|
203
|
+
fournit les outils et les preuves. Voir
|
|
206
204
|
[Programmes autorisés](#programmes-autorisés).
|
|
207
205
|
|
|
208
206
|
| | Couche 1 | Couche 2 |
|
|
209
207
|
| --- | --- | --- |
|
|
210
|
-
| Unité de travail | Un task contract |
|
|
211
|
-
| Qui le pilote | Un humain dans une session |
|
|
212
|
-
| Autorité | Plan, contract, review, checks | Ce qui précède, plus
|
|
213
|
-
| Point d'entrée | `repo-harness init` | `repo-harness
|
|
214
|
-
| Condition d'arrêt | Closeout de la tâche |
|
|
208
|
+
| Unité de travail | Un task contract | Une engineer task ordonnancée |
|
|
209
|
+
| Qui le pilote | Un humain dans une session | Le Bot, via le CLI repo-harness et Herdr |
|
|
210
|
+
| Autorité | Plan, contract, review, checks | Ce qui précède, plus engineer binding et lease |
|
|
211
|
+
| Point d'entrée | `repo-harness init` | `repo-harness engineer acquire-next` |
|
|
212
|
+
| Condition d'arrêt | Closeout de la tâche | Closeout de la tâche ou lease perdu |
|
|
215
213
|
|
|
216
214
|
La couche 2 ne remplace pas la couche 1 : chaque étape d'un programme se
|
|
217
215
|
projette toujours dans les mêmes artifacts de plan, contract et review qu'un
|
|
@@ -224,8 +222,7 @@ humain aurait écrits.
|
|
|
224
222
|
| **Sessions file-backed** | Les plans, contracts, checks et handoffs vivent dans le dépôt, si bien qu'une nouvelle session reprend à partir des artifacts plutôt que d'un chat thread |
|
|
225
223
|
| **Runtime de hooks typés** | Huit routes managées partagées, plus trois routes de delegation réservées à Codex, chacune liée à exactement un typed handler in-process, avec des guards fail-closed à la frontière d'édition |
|
|
226
224
|
| **Plan → Contract → Review** | Un seul lifecycle, du plan approuvé au contract projeté, en passant par le worktree isolé et l'evidence structurée, jusqu'à un closeout prêt pour la review |
|
|
227
|
-
| **Programmes
|
|
228
|
-
| **Controller non supervisé borné** | Une boucle de dispatch Engineer sous des caps stricts d'étapes, de durée et de retries, qui réserve du budget avant chaque tentative |
|
|
225
|
+
| **Programmes ordonnancés par le Bot** | Programmes de collaboration avec des task offers d'engineer et des leases renouvelables ; le Bot ordonnance, repo-harness fournit les outils et les preuves |
|
|
229
226
|
| **Chargement de contexte progressif** | Un root context stable d'environ 12 Ko, plus des capability contracts d'environ 1 Ko chargés uniquement pour les fichiers réellement touchés |
|
|
230
227
|
| **Intégration CodeGraph** | Requêtes structurelles (callers, callees, définitions) résolues depuis un index pré-construit, au lieu de passes grep-and-read répétées |
|
|
231
228
|
| **MCP planner sidecar** | ChatGPT lit l'état réel du dépôt et écrit les artifacts PRD/Sprint/Goal ; Codex les exécute, sans accès en écriture par défaut au source-code |
|
|
@@ -304,9 +301,9 @@ flowchart TD
|
|
|
304
301
|
PreHooks -->|blocked| ScopeFix["Fix plan, contract, worktree, or scope"]
|
|
305
302
|
ScopeFix --> Implement
|
|
306
303
|
PreHooks -->|allowed| Changes["Code, docs, tests, or config changes"]
|
|
307
|
-
Changes --> PostHooks["Post-edit and post-bash hooks<br/>trace,
|
|
308
|
-
PostHooks -->
|
|
309
|
-
|
|
304
|
+
Changes --> PostHooks["Post-edit and post-bash hooks<br/>trace, handoff, check evidence"]
|
|
305
|
+
PostHooks --> ArchUpdate["Explicit architecture update<br/>archctx ChangeSet when responsibilities change<br/>architecture-projection apply"]
|
|
306
|
+
ArchUpdate --> Verify["Run verification<br/>tests plus repo workflow checks"]
|
|
310
307
|
|
|
311
308
|
Verify --> Checks["Structured evidence<br/>.ai/harness/evidence/events/log.jsonl<br/>.ai/harness/runs/*.json"]
|
|
312
309
|
Checks --> CheckReview["Evaluator review<br/>Waza /check -> review file"]
|
|
@@ -339,40 +336,15 @@ et [`workflow-orchestration.md`](docs/reference-configs/workflow-orchestration.m
|
|
|
339
336
|
|
|
340
337
|
## Programmes autorisés
|
|
341
338
|
|
|
342
|
-
Un programme est un travail qui survit à une session.
|
|
343
|
-
|
|
339
|
+
Un programme est un travail qui survit à une session. Le Bot l'ordonnance via
|
|
340
|
+
le CLI repo-harness et Herdr. repo-harness ne garde ni scheduler, ni budget
|
|
341
|
+
ledger, ni controller qui lui soient propres.
|
|
344
342
|
|
|
345
|
-
```bash
|
|
346
|
-
repo-harness automation grant mint # store one operator ProgramAuthorizationV2
|
|
347
|
-
repo-harness automation grant list # digests held for this repository
|
|
348
|
-
repo-harness automation budget show # the enforceable per-goal ledger
|
|
349
|
-
repo-harness automation budget repair # seal a stopped or expired run's exhaustion receipt
|
|
350
|
-
```
|
|
351
|
-
|
|
352
|
-
- **Authorization.** Un `ProgramAuthorizationV2` frappé par un opérateur vit
|
|
353
|
-
dans le gate store du harness home. Il n'existe aucun chemin de démarrage non
|
|
354
|
-
authentifié, et un programme ne dérive jamais son propre actor — l'auteur de
|
|
355
|
-
chaque enregistrement est résolu depuis `--authorization-id`.
|
|
356
|
-
- **Budget.** Les agent turns, acquisitions et runner invocations réservent le budget avant le travail. `budget repair` conserve les limites.
|
|
357
343
|
- **Lease.** Le travail retenu porte un lease renouvelable avec un intervalle de
|
|
358
344
|
renouvellement, un TTL maximum, et un ensemble fermé de sources d'evidence. Un
|
|
359
345
|
état de liveness non prouvé demande de l'attention plutôt qu'une reprise
|
|
360
346
|
silencieuse.
|
|
361
347
|
|
|
362
|
-
### Controller non supervisé
|
|
363
|
-
|
|
364
|
-
```bash
|
|
365
|
-
repo-harness automation controller start --maximum-steps 20 --maximum-duration-ms 300000
|
|
366
|
-
repo-harness automation controller step
|
|
367
|
-
repo-harness automation controller status
|
|
368
|
-
repo-harness automation controller stop
|
|
369
|
-
```
|
|
370
|
-
|
|
371
|
-
Une seule boucle de dispatch Engineer sous des caps stricts, avec un backoff
|
|
372
|
-
déterministe et un ledger borné de retries de tentative. Chaque tentative
|
|
373
|
-
réserve du budget avant d'être enregistrée, et un outcome projeté en dehors de
|
|
374
|
-
l'enum fermée ne peut pas être compté comme satisfait.
|
|
375
|
-
|
|
376
348
|
### Ordonnancement des Engineers
|
|
377
349
|
|
|
378
350
|
```bash
|
|
@@ -391,20 +363,18 @@ immuables sous le backlog schema v2 — lancez une fois
|
|
|
391
363
|
|
|
392
364
|
Campaign execution moved to the existing Bot skills on 2026-10-04. Use [repo-harness](SKILL.md) to dispatch and collect work through Herdr/OAR. Use [repo-harness-product](assets/skills/repo-harness-product/SKILL.md) for planning and [repo-harness-check](assets/skill-commands/repo-harness-check/SKILL.md) for scope and verification. repo-harness has no campaign runtime.
|
|
393
365
|
|
|
394
|
-
### Refactor
|
|
366
|
+
### Refactor recommendations
|
|
395
367
|
|
|
396
368
|
```bash
|
|
397
|
-
repo-harness refactor
|
|
398
|
-
repo-harness refactor
|
|
399
|
-
repo-harness refactor verify-candidate
|
|
400
|
-
repo-harness refactor board
|
|
369
|
+
repo-harness refactor recommendations # measured suggestions with evidence
|
|
370
|
+
repo-harness refactor decide <id> accept|defer|reject --reason "<text>"
|
|
401
371
|
```
|
|
402
372
|
|
|
403
|
-
|
|
404
|
-
|
|
405
|
-
|
|
406
|
-
|
|
407
|
-
|
|
373
|
+
ArchContext mesure des problèmes structurels comme les cycles et les
|
|
374
|
+
violations de direction. repo-harness montre chaque suggestion avec ses
|
|
375
|
+
métriques et les statistiques de ses modules, et enregistre la décision de
|
|
376
|
+
l'utilisateur. Il n'exécute jamais de refactor ; une suggestion acceptée suit le
|
|
377
|
+
flux normal de plan et de pull request.
|
|
408
378
|
|
|
409
379
|
### Plan de collaboration
|
|
410
380
|
|
|
@@ -444,6 +414,8 @@ repo-harness review status --contract tasks/contracts/<task>.contract.md
|
|
|
444
414
|
repo-harness review close --contract tasks/contracts/<task>.contract.md
|
|
445
415
|
```
|
|
446
416
|
|
|
417
|
+
Dans l'application de bureau Claude sans Herdr, ajoutez `--launch app-terminal` et omettez `--herdr-endpoint`. Le premier appel affiche la commande fixe du host. Claude l'exécute une fois dans un onglet Terminal de l'application, puis relance la même ronde. OAR, Seatbelt et le Receipt restent identiques.
|
|
418
|
+
|
|
447
419
|
Un task-agent fleet `deep-reasoner` persiste dans un linked checkout dédié et la session Herdr explicitement adressée, pour trois rounds au maximum. Le harness opposé à celui du owner est choisi par défaut ; `--harness` permet un choix explicite. Seul un executable absent avant le démarrage permet un fallback annoncé. Le Result fichier passe la validation de domaine puis le writer/verifier Receipt generic-review ; close revérifie le Receipt. History reste une observation. Claude utilise une allowlist de fichiers résultat et des fingerprints ; isolation OS de production et ingestion complète des grands packets restent non vérifiées.
|
|
448
420
|
|
|
449
421
|
## Hooks
|
|
@@ -579,11 +551,12 @@ bornée pendant que le CLI et les hooks possèdent l'exécution.
|
|
|
579
551
|
| `repo-harness-product` | Modes PRD, Sprint et Goal pour le product planning upper-layer |
|
|
580
552
|
| `repo-harness-check` | Planifier le travail. Revoir le plan. Vérifier les résultats enregistrés. |
|
|
581
553
|
| `repo-harness-ship` | Valide les worktrees terminés, push les branches et ouvre les PRs |
|
|
582
|
-
| `repo-harness-architecture` |
|
|
554
|
+
| `repo-harness-architecture` | Modèle d'architecture, docs projetés et diagrammes sans rafraîchissement complet du harness |
|
|
583
555
|
| `repo-harness-cross-review` | Generic acceptance through persistent fleet deep-reasoner task-agent + Herdr review; direct advisory runtime retired |
|
|
584
556
|
| `repo-harness-chatgpt` | Consults Oracle browser/GPT Pro, setup du Connecteur MCP et bridge handoff ; setup explicite uniquement |
|
|
585
557
|
| `obsidian-memory` | Bot : lire ou écrire la mémoire sur demande explicite. |
|
|
586
558
|
| `repo-harness-test` | Worker : tests du code source et fixtures réelles. |
|
|
559
|
+
| `windows-python-first` | Worker : fichiers, texte et shell sous Windows natif avec Python. |
|
|
587
560
|
| `merge-gate` (externe) | Gate final exact-candidate ; repo-harness ne fournit aucun Skill merge-gate — voir [external tooling](docs/reference-configs/external-tooling.md) |
|
|
588
561
|
|
|
589
562
|
Les Bot skills définissent le périmètre, la délégation et les décisions. Les Worker skills exécutent les étapes.
|
|
@@ -591,7 +564,8 @@ Les Bot skills définissent le périmètre, la délégation et les décisions. L
|
|
|
591
564
|
Bot: `repo-harness`, `repo-harness-check`, `repo-harness-product`,
|
|
592
565
|
`repo-harness-ship`, `obsidian-memory`, `repo-harness-cross-review`,
|
|
593
566
|
`repo-harness-chatgpt`. Worker: `repo-harness-setup`,
|
|
594
|
-
`repo-harness-test`,
|
|
567
|
+
`repo-harness-test`, `repo-harness-architecture`, and
|
|
568
|
+
`windows-python-first`.
|
|
595
569
|
`obsidian-memory`, and `repo-harness-ship` exigent une invocation explicite.
|
|
596
570
|
Les tâches courantes utilisent un brief court et la description du PR. PRD, Sprint et Goal servent aux travaux produit demandés.
|
|
597
571
|
|
|
@@ -671,8 +645,8 @@ adopte la même policy.
|
|
|
671
645
|
|
|
672
646
|
## Release actuelle
|
|
673
647
|
|
|
674
|
-
- Package npm : `repo-harness@0.
|
|
675
|
-
- Generated workflow stamp : `repo-harness@0.
|
|
648
|
+
- Package npm : `repo-harness@0.22.1`
|
|
649
|
+
- Generated workflow stamp : `repo-harness@0.22.1+template@0.22.1`
|
|
676
650
|
- Dépôt GitHub : `Ancienttwo/repo-harness`
|
|
677
651
|
- Notes et historique de release : [`docs/CHANGELOG.md`](docs/CHANGELOG.md)
|
|
678
652
|
|