agentic-engineering-harness 0.5.1 → 0.6.1
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/README.md +161 -215
- package/dist/agents/config.js +2 -0
- package/dist/agents/config.js.map +1 -1
- package/dist/audit/intent.d.ts +20 -0
- package/dist/audit/intent.js +42 -0
- package/dist/audit/intent.js.map +1 -0
- package/dist/audit/run.d.ts +46 -0
- package/dist/audit/run.js +147 -0
- package/dist/audit/run.js.map +1 -0
- package/dist/cli.js +2 -1
- package/dist/cli.js.map +1 -1
- package/dist/core/assets.d.ts +13 -0
- package/dist/core/assets.js +115 -0
- package/dist/core/assets.js.map +1 -0
- package/dist/core/config.js +6 -2
- package/dist/core/config.js.map +1 -1
- package/dist/core/controlPlane.js +1 -1
- package/dist/core/controlPlane.js.map +1 -1
- package/dist/core/init.js +7 -13
- package/dist/core/init.js.map +1 -1
- package/dist/core/types.d.ts +20 -0
- package/dist/entry.js +144 -3
- package/dist/entry.js.map +1 -1
- package/dist/issues/intake.d.ts +1 -1
- package/dist/paseo/capabilities.d.ts +24 -0
- package/dist/paseo/capabilities.js +104 -0
- package/dist/paseo/capabilities.js.map +1 -0
- package/dist/paseo/context.d.ts +45 -0
- package/dist/paseo/context.js +153 -0
- package/dist/paseo/context.js.map +1 -0
- package/dist/paseo/runtime.d.ts +31 -0
- package/dist/paseo/runtime.js +158 -0
- package/dist/paseo/runtime.js.map +1 -0
- package/dist/paseo/sdk.d.ts +79 -0
- package/dist/paseo/sdk.js +165 -0
- package/dist/paseo/sdk.js.map +1 -0
- package/dist/paseo/start.d.ts +16 -2
- package/dist/paseo/start.js +58 -77
- package/dist/paseo/start.js.map +1 -1
- package/dist/spec/openspec.d.ts +28 -0
- package/dist/spec/openspec.js +179 -0
- package/dist/spec/openspec.js.map +1 -0
- package/dist/toolchain/resolve.js +3 -0
- package/dist/toolchain/resolve.js.map +1 -1
- package/dist/toolchain/setup.js +2 -0
- package/dist/toolchain/setup.js.map +1 -1
- package/dist/version.d.ts +2 -0
- package/dist/version.js +6 -0
- package/dist/version.js.map +1 -0
- package/dist/workers/agentPrompt.js +24 -29
- package/dist/workers/agentPrompt.js.map +1 -1
- package/dist/workers/paseo.d.ts +0 -1
- package/dist/workers/paseo.js +67 -21
- package/dist/workers/paseo.js.map +1 -1
- package/docs/PASEO.md +87 -4
- package/docs/PUBLISHING.md +53 -44
- package/docs/V0.5.2.md +180 -0
- package/docs/V0.6.md +148 -0
- package/package.json +87 -12
- package/presets/agents/orchestration.jsonc +37 -0
- package/schemas/project.schema.json +13 -2
- package/skills/engineering-workflow/SKILL.md +110 -86
- package/skills/openspec-authoring/SKILL.md +26 -0
- package/skills/paseo-orchestration/SKILL.md +38 -0
- package/templates/AGENTS.md +36 -17
- package/templates/agents.source.jsonc +1 -1
- package/templates/openspec-config.yaml +24 -0
- package/templates/project.yaml +11 -2
- package/templates/toolchain.yaml +11 -1
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"paseo.js","sourceRoot":"","sources":["../../src/workers/paseo.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,6BAA6B,EAAE,MAAM,0BAA0B,CAAC;AAEzE,OAAO,EAAE,mBAAmB,EAAE,MAAM,wBAAwB,CAAC;AAC7D,OAAO,EAAE,
|
|
1
|
+
{"version":3,"file":"paseo.js","sourceRoot":"","sources":["../../src/workers/paseo.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,6BAA6B,EAAE,MAAM,0BAA0B,CAAC;AAEzE,OAAO,EAAE,mBAAmB,EAAE,MAAM,wBAAwB,CAAC;AAC7D,OAAO,EAAE,uBAAuB,EAAE,MAAM,0BAA0B,CAAC;AACnE,OAAO,EAAE,yBAAyB,EAAE,uBAAuB,EAAE,MAAM,qBAAqB,CAAC;AACzF,OAAO,EAAE,aAAa,EAAE,MAAM,qBAAqB,CAAC;AACpD,OAAO,EAAE,iBAAiB,EAAE,iBAAiB,EAAE,MAAM,aAAa,CAAC;AAGnE,MAAM,OAAO,mBAAmB;IACrB,IAAI,GAAG,OAAO,CAAC;IACxB,KAAK,CAAC,MAAM,CAAC,IAAY,EAAE,OAA6B,EAAE,SAAmC;QAC3F,MAAM,EAAE,GAAG,MAAM,aAAa,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC;QAAC,IAAI,CAAC,EAAE;YAAE,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,OAAO,EAAE,sBAAsB,EAAE,CAAC;QAC9G,IAAI,SAAS,EAAE,CAAC;YAAC,MAAM,MAAM,GAAG,6BAA6B,CAAC,SAAS,EAAE,OAAO,CAAC,CAAC;YAAC,IAAI,MAAM,CAAC,MAAM;gBAAE,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,OAAO,EAAE,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC;QAAC,CAAC;QACzJ,IAAI,CAAC;YACH,MAAM,IAAI,GAAG,MAAM,uBAAuB,CAAC,IAAI,CAAC,CAAC;YACjD,IAAI,CAAC,IAAI,CAAC,UAAU;gBAAE,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,OAAO,EAAE,kBAAkB,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,IAAI,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,6EAA6E,EAAE,CAAC;YAC3L,OAAO,EAAE,EAAE,EAAE,IAAI,EAAE,OAAO,EAAE,iBAAiB,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,KAAK,IAAI,CAAC,OAAO,GAAG,CAAC,CAAC,CAAC,EAAE,4FAA4F,EAAE,CAAC;QACtL,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YAAC,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,OAAO,EAAE,kCAAkC,MAAM,CAAC,KAAK,CAAC,EAAE,EAAE,CAAC;QAAC,CAAC;IACvG,CAAC;IACD,KAAK,CAAC,KAAK,CAAC,IAAY,EAAE,MAA4B,EAAE,QAAsB,EAAE,SAAmC,IAA4B,OAAO,IAAI,CAAC,MAAM,CAAC,IAAI,EAAE,MAAM,EAAE,QAAQ,EAAE,iBAAiB,CAAC,QAAQ,EAAE,SAAS,CAAC,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC;IAC/O,KAAK,CAAC,MAAM,CAAC,IAAY,EAAE,MAA4B,EAAE,QAAsB,EAAE,OAAsB,EAAE,MAAoB,EAAE,SAAmC;QAChK,IAAI,OAAO,CAAC,EAAE,IAAI,CAAC,CAAC,SAAS,IAAI,OAAO,CAAC,YAAY,KAAK,SAAS,CAAC,YAAY,CAAC,EAAE,CAAC;YAClF,MAAM,SAAS,GAAG,MAAM,yBAAyB,CAAC,IAAI,EAAE,OAAO,CAAC,EAAE,EAAE,iBAAiB,CAAC,MAAM,CAAC,EAAE,MAAM,CAAC,aAAa,EAAE,MAAM,EAAE,cAAc,CAAC,CAAC;YAC7I,OAAO,EAAE,GAAG,OAAO,EAAE,QAAQ,EAAE,SAAS,CAAC,QAAQ,EAAE,MAAM,EAAE,SAAS,CAAC,MAAM,IAAI,OAAO,CAAC,MAAM,EAAE,MAAM,EAAE,CAAC,OAAO,CAAC,MAAM,EAAE,SAAS,CAAC,MAAM,CAAC,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC;QACzK,CAAC;QACD,OAAO,IAAI,CAAC,MAAM,CAAC,IAAI,EAAE,MAAM,EAAE,QAAQ,EAAE,GAAG,iBAAiB,CAAC,QAAQ,EAAE,SAAS,CAAC,OAAO,iBAAiB,CAAC,MAAM,CAAC,EAAE,EAAE,SAAS,CAAC,CAAC;IACrI,CAAC;IACO,KAAK,CAAC,MAAM,CAAC,IAAY,EAAE,MAA4B,EAAE,QAAsB,EAAE,MAAc,EAAE,SAAmC;QAC1I,MAAM,MAAM,GAAG,MAAM,CAAC,aAAa,EAAE,MAAM,CAAC;QAC5C,MAAM,QAAQ,GAAG,SAAS,EAAE,aAAa,IAAI,MAAM,EAAE,QAAQ,IAAI,UAAU,CAAC;QAC5E,MAAM,KAAK,GAAG,SAAS,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,cAAc,KAAK,OAAO,CAAC,CAAC,CAAC,SAAS,CAAC,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,MAAM,EAAE,KAAK,CAAC;QAC3H,MAAM,YAAY,GAAG,SAAS,EAAE,YAAY,IAAI,QAAQ,CAAC;QACzD,MAAM,KAAK,GAAG,GAAG,MAAM,EAAE,WAAW,IAAI,KAAK,IAAI,QAAQ,CAAC,IAAI,CAAC,EAAE,IAAI,YAAY,EAAE,CAAC;QACpF,MAAM,WAAW,GAAG,MAAM,mBAAmB,CAAC,IAAI,EAAE,MAAM,EAAE,QAAQ,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QAC9E,MAAM,MAAM,GAA2B;YACrC,aAAa,EAAE,MAAM,CAAC,OAAO,CAAC,IAAI;YAClC,UAAU,EAAE,QAAQ;YACpB,UAAU,EAAE,QAAQ,CAAC,IAAI,CAAC,EAAE;YAC5B,UAAU,EAAE,YAAY;SACzB,CAAC;QACF,IAAI,SAAS,EAAE,OAAO;YAAE,MAAM,CAAC,aAAa,CAAC,GAAG,SAAS,CAAC,OAAO,CAAC;QAClE,MAAM,QAAQ,GAAG,MAAM,uBAAuB,CAAC,IAAI,EAAE;YACnD,GAAG,EAAE,IAAI;YACT,KAAK;YACL,QAAQ;YACR,KAAK;YACL,WAAW;YACX,MAAM;YACN,MAAM;YACN,cAAc,EAAE,MAAM,EAAE,cAAc,IAAI,IAAI;SAC/C,CAAC,CAAC;QACH,OAAO;YACL,EAAE,EAAE,QAAQ,CAAC,EAAE;YACf,QAAQ;YACR,KAAK;YACL,YAAY,EAAE,SAAS,EAAE,YAAY;YACrC,WAAW,EAAE,SAAS,EAAE,WAAW;YACnC,OAAO,EAAE,SAAS,EAAE,WAAW;YAC/B,OAAO,EAAE,SAAS,EAAE,OAAO;YAC3B,QAAQ,EAAE,QAAQ,CAAC,QAAQ;YAC3B,MAAM,EAAE,QAAQ,CAAC,MAAM;YACvB,MAAM,EAAE,QAAQ,CAAC,MAAM;SACxB,CAAC;IACJ,CAAC;CACF"}
|
package/docs/PASEO.md
CHANGED
|
@@ -1,9 +1,92 @@
|
|
|
1
1
|
# Paseo Integration
|
|
2
2
|
|
|
3
|
-
Paseo is
|
|
3
|
+
Paseo is AEH's default interactive orchestration surface. The integration separates **Paseo communication/session lifecycle** from **AEH deterministic workflow ownership**.
|
|
4
4
|
|
|
5
|
-
|
|
5
|
+
## SDK-first control plane
|
|
6
6
|
|
|
7
|
-
|
|
7
|
+
AEH uses the official TypeScript SDK, `@getpaseo/client`, as the primary control surface for agent creation, follow-up turns, status lookup and directory queries. The SDK is resolved directly when available and otherwise from the managed `@getpaseo/cli` installation. Paseo CLI already ships the matching client package, so AEH does not independently pin a second Paseo SDK version.
|
|
8
8
|
|
|
9
|
-
|
|
9
|
+
The CLI remains responsible for daemon bootstrap/recovery and is retained as a compatibility fallback when the SDK cannot be resolved or connected:
|
|
10
|
+
|
|
11
|
+
```text
|
|
12
|
+
AEH
|
|
13
|
+
├── daemon/bootstrap/recovery -> Paseo CLI
|
|
14
|
+
└── normal agent lifecycle -> @getpaseo/client
|
|
15
|
+
```
|
|
16
|
+
|
|
17
|
+
Set `AEH_PASEO_FORCE_CLI=1` only when the compatibility path is deliberately required. `PASEO_DAEMON_URL` overrides the default SDK endpoint `ws://127.0.0.1:6767/ws`; `PASEO_DAEMON_PASSWORD` supplies daemon authentication when configured.
|
|
18
|
+
|
|
19
|
+
A system-prompt-only idle agent is an SDK-only invariant. If the SDK is unavailable, AEH refuses to degrade that lead creation into a CLI user turn because doing so would expose the bootstrap as conversational input.
|
|
20
|
+
|
|
21
|
+
## Conversational lead
|
|
22
|
+
|
|
23
|
+
`aeh start` creates the lead as an idle Paseo agent. The AEH bootstrap is passed through the SDK's `systemPrompt`; it is no longer sent as the first user message and there is no synthetic `AEH READY` turn.
|
|
24
|
+
|
|
25
|
+
The bootstrap is intentionally thin. `AGENTS.md`, `.harness/skills/engineering-workflow/SKILL.md`, and the resolved AEH agent topology are authoritative for roles, charters, permissions and delegation. This avoids duplicating a role map inside every Paseo conversation and prevents prompt/configuration drift.
|
|
26
|
+
|
|
27
|
+
When the lead is running inside Paseo and Paseo exposes its orchestration tools, it may still use the native/MCP conversational surface and `/paseo-handoff`. Deterministic Harness-owned work, however, is created externally by AEH through the SDK.
|
|
28
|
+
|
|
29
|
+
## Independent AEH agents
|
|
30
|
+
|
|
31
|
+
AEH-managed workers are created without the Paseo SDK `parent` option. A worker may be placed in a delivery workspace, but workspace placement does not establish parentage. This makes each worker a top-level Paseo agent rather than a child whose lifecycle belongs to the current lead conversation.
|
|
32
|
+
|
|
33
|
+
Workflow ownership is represented by labels instead:
|
|
34
|
+
|
|
35
|
+
```text
|
|
36
|
+
aeh.project=<project>
|
|
37
|
+
aeh.kind=lead|worker
|
|
38
|
+
aeh.role=<logical-agent>
|
|
39
|
+
aeh.task=<task-id> # workers
|
|
40
|
+
aeh.profile=<profile> # when selected
|
|
41
|
+
aeh.generation=<n> # leads
|
|
42
|
+
```
|
|
43
|
+
|
|
44
|
+
The shared Paseo runtime exposes list/inspect primitives over those labels. This lets AEH determine which logical agent is active for a task without scraping assistant prose or relying on parent/child nesting. The same labels survive lead context rotation, so a fresh lead can correlate existing workers with durable AEH run state.
|
|
45
|
+
|
|
46
|
+
### Observe active agents
|
|
47
|
+
|
|
48
|
+
Use the top-level AEH command to inspect the live Paseo directory. The project label is applied automatically; filters are additive:
|
|
49
|
+
|
|
50
|
+
```bash
|
|
51
|
+
aeh paseo agents
|
|
52
|
+
aeh paseo agents --status working
|
|
53
|
+
aeh paseo agents --kind worker --role backend-implementer
|
|
54
|
+
aeh paseo agents --task TASK-123 --json
|
|
55
|
+
```
|
|
56
|
+
|
|
57
|
+
The tabular view reports status, logical role, task, kind, stable Paseo agent ID and title. `--json` returns the same normalized fields plus the complete AEH label set. This is the preferred way to answer which AEH agent is currently working without depending on Paseo parent/subagent nesting.
|
|
58
|
+
|
|
59
|
+
## Runtime consolidation
|
|
60
|
+
|
|
61
|
+
Lead startup, `PaseoWorkerExecutor`, and generic `agentPrompt` execution all use the same managed Paseo runtime. That runtime owns SDK-first create/run/probe/list behavior and the CLI fallback. Individual worker paths should not add new hand-written `paseo run/send/wait/logs` loops.
|
|
62
|
+
|
|
63
|
+
Structured output is passed through the SDK `outputSchema` field. The compatibility CLI path continues to negotiate `--output-schema` and background capabilities dynamically.
|
|
64
|
+
|
|
65
|
+
## Session policy
|
|
66
|
+
|
|
67
|
+
Default:
|
|
68
|
+
|
|
69
|
+
```yaml
|
|
70
|
+
orchestration:
|
|
71
|
+
interactive:
|
|
72
|
+
sessionPolicy: fresh-on-start
|
|
73
|
+
```
|
|
74
|
+
|
|
75
|
+
```bash
|
|
76
|
+
aeh start # fresh idle lead
|
|
77
|
+
aeh start --resume # explicit compatible reuse
|
|
78
|
+
```
|
|
79
|
+
|
|
80
|
+
Workspaces and durable AEH state remain reusable even though normal conversational context starts clean.
|
|
81
|
+
|
|
82
|
+
## Context rotation
|
|
83
|
+
|
|
84
|
+
Default pressure policy is 70/80/90 percent. `aeh context guard` consumes a context ratio only when Paseo exposes a usable field; AEH does not guess.
|
|
85
|
+
|
|
86
|
+
At the handoff threshold it writes a deterministic `.harness/paseo/handoffs/*.json` artifact. From a managed Paseo lead it also creates the replacement lead automatically and bootstraps it from the handoff artifact and referenced sealed/run/audit/delivery state. Workers remain independent top-level agents associated through AEH labels and durable task/run state rather than lead parentage.
|
|
87
|
+
|
|
88
|
+
## Trust boundary
|
|
89
|
+
|
|
90
|
+
Paseo owns communication, process and session lifecycle. It is not normative engineering truth. AEH TaskContracts, seals, deterministic reports, evidence graphs and quality gates continue to decide acceptance.
|
|
91
|
+
|
|
92
|
+
For stronger direct process isolation configure Podman sandboxing where appropriate; Paseo orchestration and worker sandboxing remain separate policy axes.
|
package/docs/PUBLISHING.md
CHANGED
|
@@ -2,41 +2,62 @@
|
|
|
2
2
|
|
|
3
3
|
The package name is `agentic-engineering-harness` and the public CLI commands are `aeh` and `engineering-harness`.
|
|
4
4
|
|
|
5
|
-
|
|
5
|
+
`package.json` is the single source of truth for the AEH version. Runtime CLI version output imports that package metadata; source files and CI must not maintain separate hard-coded version strings.
|
|
6
6
|
|
|
7
7
|
## Preflight
|
|
8
8
|
|
|
9
|
-
|
|
9
|
+
Every release candidate must pass:
|
|
10
10
|
|
|
11
11
|
```bash
|
|
12
12
|
npm run release:check
|
|
13
13
|
```
|
|
14
14
|
|
|
15
|
-
This
|
|
15
|
+
This runs typecheck, the complete test suite, build and `npm pack --dry-run`.
|
|
16
16
|
|
|
17
|
-
|
|
17
|
+
## Automatic releases from `main`
|
|
18
18
|
|
|
19
|
-
|
|
20
|
-
|
|
19
|
+
`.github/workflows/publish.yml` is the single npm publishing workflow. A push to `main` starts an idempotent release pipeline unless the repository variable below is set:
|
|
20
|
+
|
|
21
|
+
```text
|
|
22
|
+
AEH_AUTO_PUBLISH=false
|
|
21
23
|
```
|
|
22
24
|
|
|
23
|
-
|
|
25
|
+
The workflow performs the following steps:
|
|
24
26
|
|
|
25
|
-
|
|
27
|
+
1. installs dependencies with `npm ci`;
|
|
28
|
+
2. checks whether the current `package.json` version is already present on npm;
|
|
29
|
+
3. if the current version is unpublished, it publishes that exact version first;
|
|
30
|
+
4. otherwise it derives the next semantic version from commits since the latest `v*` tag:
|
|
31
|
+
- a breaking Conventional Commit (`type!:` or `BREAKING CHANGE:`) -> major;
|
|
32
|
+
- `feat:` -> minor;
|
|
33
|
+
- every other change -> patch;
|
|
34
|
+
5. synchronizes `package.json` and `package-lock.json` with `npm version --no-git-tag-version`;
|
|
35
|
+
6. runs `npm run release:check` on the exact candidate;
|
|
36
|
+
7. commits the version metadata as `chore(release): vX.Y.Z [skip ci]` and creates the matching Git tag;
|
|
37
|
+
8. publishes the package to npm with provenance;
|
|
38
|
+
9. creates the GitHub Release for the tag.
|
|
26
39
|
|
|
27
|
-
|
|
40
|
+
The release commit/tag is pushed with GitHub's repository token. GitHub does not recursively trigger ordinary push workflows for pushes created with that `GITHUB_TOKEN`, so the version commit does not create an infinite publish loop.
|
|
28
41
|
|
|
29
|
-
|
|
42
|
+
The repository must allow the workflow identity to write the release metadata commit/tag. If branch rules forbid direct writes to `main`, grant the GitHub Actions identity the appropriate bypass/write permission or set `AEH_AUTO_PUBLISH=false` until the repository rule is adjusted. Source validation remains independent of publication.
|
|
30
43
|
|
|
31
|
-
|
|
32
|
-
|
|
33
|
-
npm
|
|
34
|
-
|
|
44
|
+
## Manual release control
|
|
45
|
+
|
|
46
|
+
`publish-npm` also supports `workflow_dispatch`. The `bump` input can be:
|
|
47
|
+
|
|
48
|
+
```text
|
|
49
|
+
auto # Conventional Commit-derived bump
|
|
50
|
+
current # publish current version only if it is not already published
|
|
51
|
+
patch
|
|
52
|
+
minor
|
|
53
|
+
major
|
|
35
54
|
```
|
|
36
55
|
|
|
37
|
-
|
|
56
|
+
Manual dispatch is useful for retrying an external npm/OIDC failure or deliberately overriding the automatic bump classification. A `current` retry is also able to recreate a missing GitHub Release when the npm version and matching Git tag already exist; it will not fabricate a missing tag for an already-published package.
|
|
38
57
|
|
|
39
|
-
|
|
58
|
+
## npm authentication
|
|
59
|
+
|
|
60
|
+
The preferred steady-state path is npm Trusted Publishing with GitHub Actions OIDC. Configure the npm package Trusted Publisher with:
|
|
40
61
|
|
|
41
62
|
```text
|
|
42
63
|
Provider: GitHub Actions
|
|
@@ -46,39 +67,21 @@ Workflow filename: publish.yml
|
|
|
46
67
|
Allowed action: npm publish
|
|
47
68
|
```
|
|
48
69
|
|
|
49
|
-
The workflow
|
|
50
|
-
|
|
51
|
-
For the strongest steady-state posture, after the OIDC flow has been proven once, disallow traditional publish tokens for the package and retain 2FA on the maintainer account.
|
|
70
|
+
The workflow grants `id-token: write`, which is required for OIDC. Modern npm clients can exchange the GitHub OIDC identity for short-lived publish authorization, avoiding a long-lived npm write token.
|
|
52
71
|
|
|
53
|
-
|
|
72
|
+
For bootstrap or compatibility, the workflow also accepts an optional GitHub Actions secret named `NPM_TOKEN`. If present, it is exported only for the `npm publish` step. Once Trusted Publishing is verified, prefer removing the long-lived token.
|
|
54
73
|
|
|
55
|
-
|
|
56
|
-
2. Ensure the CLI dispatcher reports the same version.
|
|
57
|
-
3. Merge only after CI `release:check` passes.
|
|
58
|
-
4. Create/publish a GitHub Release tagged exactly:
|
|
74
|
+
If the package has never been published and npm does not permit Trusted Publisher configuration before first publication, perform one maintainer-authenticated bootstrap publish, then configure the Trusted Publisher above. The automatic workflow will subsequently see that version as published and continue normal semantic versioning.
|
|
59
75
|
|
|
60
|
-
|
|
61
|
-
v<package.json version>
|
|
62
|
-
```
|
|
63
|
-
|
|
64
|
-
For example:
|
|
65
|
-
|
|
66
|
-
```text
|
|
67
|
-
v0.4.16
|
|
68
|
-
```
|
|
76
|
+
## Version policy
|
|
69
77
|
|
|
70
|
-
|
|
71
|
-
- check out the release commit;
|
|
72
|
-
- use Node 24 on a GitHub-hosted runner;
|
|
73
|
-
- verify that the release tag exactly matches `package.json`;
|
|
74
|
-
- run `npm run release:check` again;
|
|
75
|
-
- execute `npm publish` using npm Trusted Publishing/OIDC.
|
|
78
|
+
The repository currently starts this release line at `0.6.1`. After that, normal merges do not require a human to edit the version manually. The release workflow owns the release metadata bump.
|
|
76
79
|
|
|
77
|
-
If the
|
|
80
|
+
If a PR intentionally changes the package version to a version that is not yet on npm, that repository version wins: the next `main` publication ships it before any further automatic increment. This makes explicit release corrections and recovery deterministic.
|
|
78
81
|
|
|
79
82
|
## What enters the npm tarball
|
|
80
83
|
|
|
81
|
-
The `files` allowlist in `package.json` publishes
|
|
84
|
+
The `files` allowlist in `package.json` publishes:
|
|
82
85
|
|
|
83
86
|
```text
|
|
84
87
|
dist/
|
|
@@ -92,7 +95,7 @@ docs/
|
|
|
92
95
|
|
|
93
96
|
plus npm-required package metadata such as `package.json`, README and LICENSE.
|
|
94
97
|
|
|
95
|
-
|
|
98
|
+
Packaged `skills/` and core `policies/` are runtime control-plane assets. `aeh init`, `aeh setup`, and `aeh start` reconcile those package assets into the consumer repository's `.harness` directory. `.harness/managed-assets.json` is versioned project state: it records hashes so missing or untouched files can be restored/upgraded, retired untouched assets can be removed, and project-local modifications remain protected as overrides.
|
|
96
99
|
|
|
97
100
|
## Consumer installation
|
|
98
101
|
|
|
@@ -103,8 +106,14 @@ npm install --save-dev agentic-engineering-harness
|
|
|
103
106
|
npm exec aeh -- init --setup
|
|
104
107
|
```
|
|
105
108
|
|
|
106
|
-
|
|
109
|
+
After the repository has been initialized, a normal:
|
|
110
|
+
|
|
111
|
+
```bash
|
|
112
|
+
npm exec aeh -- start
|
|
113
|
+
```
|
|
114
|
+
|
|
115
|
+
reconciles managed Harness assets before loading the agent topology and starting Paseo.
|
|
107
116
|
|
|
108
117
|
## Failure policy
|
|
109
118
|
|
|
110
|
-
|
|
119
|
+
A registry/OIDC/permission failure is an external delivery failure, not a reason to rewrite validated engineering history. The release workflow is retry-safe: if the version commit/tag exists but npm publication failed, a manual rerun with `current` will attempt the same unpublished version rather than incrementing it again. If npm publication succeeded but GitHub Release creation failed, a `current` rerun can repair the release from the existing matching tag.
|
package/docs/V0.5.2.md
ADDED
|
@@ -0,0 +1,180 @@
|
|
|
1
|
+
# v0.5.2 — Harness-governed engineering audits
|
|
2
|
+
|
|
3
|
+
v0.5.2 closes the read-only escape that existed in v0.5.1. A request no longer bypasses AEH merely because it does not mutate files. The interactive entrypoint now distinguishes a purely informational question from a read-only engineering operation.
|
|
4
|
+
|
|
5
|
+
## Intent model
|
|
6
|
+
|
|
7
|
+
Every conversational request is classified at the top level as:
|
|
8
|
+
|
|
9
|
+
```text
|
|
10
|
+
prompt
|
|
11
|
+
|
|
|
12
|
+
v
|
|
13
|
+
engineering intent
|
|
14
|
+
|-- INFORMATIONAL -> direct answer, no mutation
|
|
15
|
+
|-- AUDIT -> Harness read-only audit pipeline
|
|
16
|
+
`-- CHANGE -> deterministic QUICK | SPEC triage
|
|
17
|
+
```
|
|
18
|
+
|
|
19
|
+
`AUDIT` is deliberately not a subtype of QUICK. A repository-wide review is valid audit scope even though it would be invalid QUICK scope, and it does not need artificial SDD artifacts merely to run a code review.
|
|
20
|
+
|
|
21
|
+
Examples:
|
|
22
|
+
|
|
23
|
+
```text
|
|
24
|
+
"What does src/core/run.ts do?"
|
|
25
|
+
-> INFORMATIONAL
|
|
26
|
+
|
|
27
|
+
"review the repo and validate the code for improvements"
|
|
28
|
+
-> AUDIT
|
|
29
|
+
|
|
30
|
+
"find bugs and security risks"
|
|
31
|
+
-> AUDIT
|
|
32
|
+
|
|
33
|
+
"review this module and fix every bug you find"
|
|
34
|
+
-> CHANGE -> QUICK | SPEC
|
|
35
|
+
```
|
|
36
|
+
|
|
37
|
+
The public classifier is:
|
|
38
|
+
|
|
39
|
+
```bash
|
|
40
|
+
aeh intent "review the repo and validate the code for improvements"
|
|
41
|
+
```
|
|
42
|
+
|
|
43
|
+
## Audit workflow
|
|
44
|
+
|
|
45
|
+
Run an audit explicitly with:
|
|
46
|
+
|
|
47
|
+
```bash
|
|
48
|
+
aeh audit "review the repo and validate the code for improvements"
|
|
49
|
+
```
|
|
50
|
+
|
|
51
|
+
Optional hints include repeated `--file`, `--domain`, `--reviewer` and `--risk low|medium|high` values.
|
|
52
|
+
|
|
53
|
+
The pipeline is:
|
|
54
|
+
|
|
55
|
+
```text
|
|
56
|
+
AUDIT request
|
|
57
|
+
|
|
|
58
|
+
v
|
|
59
|
+
worktree checkpoint
|
|
60
|
+
|
|
|
61
|
+
v
|
|
62
|
+
transient TaskContract + SHA-256 seal
|
|
63
|
+
|
|
|
64
|
+
v
|
|
65
|
+
hard-frozen control plane
|
|
66
|
+
|
|
|
67
|
+
v
|
|
68
|
+
configured deterministic validators
|
|
69
|
+
|
|
|
70
|
+
v
|
|
71
|
+
failure classification
|
|
72
|
+
|
|
|
73
|
+
v
|
|
74
|
+
read-only reviewer wave
|
|
75
|
+
|
|
|
76
|
+
v
|
|
77
|
+
finding normalization + deduplication
|
|
78
|
+
|
|
|
79
|
+
v
|
|
80
|
+
severity counts + DebtScore + Quality Gate evaluation
|
|
81
|
+
|
|
|
82
|
+
v
|
|
83
|
+
AuditReport
|
|
84
|
+
```
|
|
85
|
+
|
|
86
|
+
Reviewer sessions are clamped to `write=deny`, `gitWrite=deny` and `delegate=deny`. The worktree is also checkpointed before the audit and rolled back afterward, so an accidental tracked mutation by a reviewer or tool does not survive. Any dirty paths that existed before the audit are restored to their exact prior contents.
|
|
87
|
+
|
|
88
|
+
The transient contract and seal make the audit compatible with the same direct, Paseo and supported Podman execution boundaries used elsewhere in AEH.
|
|
89
|
+
|
|
90
|
+
## Deterministic failures are evidence
|
|
91
|
+
|
|
92
|
+
The lead must not silently reinterpret a failing validator as harmless environment noise. Audit reports classify failed checks when possible as:
|
|
93
|
+
|
|
94
|
+
```text
|
|
95
|
+
ASSERTION_FAILURE
|
|
96
|
+
ENVIRONMENT_FAILURE
|
|
97
|
+
SANDBOX_DENIAL
|
|
98
|
+
MISSING_DEPENDENCY
|
|
99
|
+
TOOL_FAILURE
|
|
100
|
+
```
|
|
101
|
+
|
|
102
|
+
For example:
|
|
103
|
+
|
|
104
|
+
```text
|
|
105
|
+
spawnSync git EPERM: operation not permitted
|
|
106
|
+
-> SANDBOX_DENIAL
|
|
107
|
+
```
|
|
108
|
+
|
|
109
|
+
That classification does not pretend the affected test passed. The audit remains `DEGRADED` and `productionSafe=false` when a deterministic failure prevents establishing correctness.
|
|
110
|
+
|
|
111
|
+
## AuditReport
|
|
112
|
+
|
|
113
|
+
Reports are machine-readable local artifacts:
|
|
114
|
+
|
|
115
|
+
```text
|
|
116
|
+
.harness/audits/<auditId>.json
|
|
117
|
+
.harness/audits/latest.json
|
|
118
|
+
```
|
|
119
|
+
|
|
120
|
+
An AuditReport contains:
|
|
121
|
+
|
|
122
|
+
- frozen controller SHA;
|
|
123
|
+
- audited commit/base ref and pre-existing dirty paths;
|
|
124
|
+
- deterministic checks plus failure classes;
|
|
125
|
+
- reviewer sessions;
|
|
126
|
+
- normalized/deduplicated findings;
|
|
127
|
+
- severity counts;
|
|
128
|
+
- integer DebtPoints and displayed DebtScore;
|
|
129
|
+
- Final Quality Gate evaluation;
|
|
130
|
+
- conservative `productionSafe` status;
|
|
131
|
+
- any paths restored by the read-only rollback boundary.
|
|
132
|
+
|
|
133
|
+
Audit artifacts are generated state and are gitignored by `aeh init`.
|
|
134
|
+
|
|
135
|
+
## Audit to remediation
|
|
136
|
+
|
|
137
|
+
AUDIT itself does not edit files or auto-remediate. If the user later says:
|
|
138
|
+
|
|
139
|
+
```text
|
|
140
|
+
fix these
|
|
141
|
+
fix all findings
|
|
142
|
+
implement the audit recommendations
|
|
143
|
+
```
|
|
144
|
+
|
|
145
|
+
the persistent lead reads `.harness/audits/latest.json` and the referenced report, reuses the normalized findings as evidence for a new `CHANGE`, and performs normal deterministic change triage.
|
|
146
|
+
|
|
147
|
+
```text
|
|
148
|
+
AuditReport
|
|
149
|
+
|
|
|
150
|
+
v
|
|
151
|
+
new CHANGE request
|
|
152
|
+
|
|
|
153
|
+
v
|
|
154
|
+
QUICK | SPEC
|
|
155
|
+
|
|
|
156
|
+
v
|
|
157
|
+
sealed implementation workflow
|
|
158
|
+
```
|
|
159
|
+
|
|
160
|
+
The audit result does not override safety policy. Security, architecture, schema, public API, cross-module or otherwise unsafe findings still force SPEC when normal change triage requires it.
|
|
161
|
+
|
|
162
|
+
## Paseo behavior
|
|
163
|
+
|
|
164
|
+
`aeh start` now writes bootstrap version 2. Existing v0.5.1 lead sessions are intentionally incompatible and are recreated/replaced on the next start so the old broad read-only bypass cannot persist.
|
|
165
|
+
|
|
166
|
+
The standing interactive invariant is now:
|
|
167
|
+
|
|
168
|
+
> Every engineering operation must enter through the Harness, whether read-only or mutating. Only purely informational questions may bypass AEH.
|
|
169
|
+
|
|
170
|
+
Consequently this prompt:
|
|
171
|
+
|
|
172
|
+
```text
|
|
173
|
+
review the repo and validate the code for improvements
|
|
174
|
+
```
|
|
175
|
+
|
|
176
|
+
must route to `aeh audit`; the lead must not answer that it skipped AEH because the work is read-only.
|
|
177
|
+
|
|
178
|
+
## Control-plane freeze
|
|
179
|
+
|
|
180
|
+
Self-hosted AEH snapshots now explicitly include `src/audit` and `src/paseo` in the controller trust boundary. Audit policy and the interactive routing code therefore cannot modify themselves and take effect inside the same active audit/run; changes govern only a later invocation after validation.
|
package/docs/V0.6.md
ADDED
|
@@ -0,0 +1,148 @@
|
|
|
1
|
+
# v0.6 — Orchestration & Context Architecture
|
|
2
|
+
|
|
3
|
+
v0.6 changes the interactive model from a persistent lead that personally performs setup, discovery and specification authoring into a thin semantic orchestrator backed by bounded agents and durable Harness state.
|
|
4
|
+
|
|
5
|
+
## Thin lead
|
|
6
|
+
|
|
7
|
+
The interactive lead owns user intent, high-level routing, true ambiguity and final semantic acceptance. It delegates operational work:
|
|
8
|
+
|
|
9
|
+
```text
|
|
10
|
+
User
|
|
11
|
+
|
|
|
12
|
+
v
|
|
13
|
+
AEH Lead
|
|
14
|
+
|-- explorer repository discovery
|
|
15
|
+
|-- environment-manager doctor/setup/Paseo recovery
|
|
16
|
+
|-- planner decomposition + triage evidence
|
|
17
|
+
|-- spec-manager OpenSpec authoring
|
|
18
|
+
`-- AEH implementation/validation/review/delivery
|
|
19
|
+
```
|
|
20
|
+
|
|
21
|
+
The lead should receive compact structured outcomes instead of retaining raw shell logs, full source files or long environment-repair transcripts.
|
|
22
|
+
|
|
23
|
+
## Paseo-native orchestration first
|
|
24
|
+
|
|
25
|
+
When the current Paseo agent exposes its orchestration tools, the lead prefers `create_agent`, `send_agent_prompt`, `get_agent_status`, `get_agent_activity` and lifecycle controls, and uses `/paseo-handoff` for responsibility transfer. `/paseo-committee` and `/paseo-advisor` may provide advisory reasoning but never override deterministic AEH gates.
|
|
26
|
+
|
|
27
|
+
AEH still contains a CLI-based Paseo adapter because an external Node control-plane process cannot invoke tools local to the conversational agent. That adapter is now a compatibility fallback rather than the lead's normal orchestration mechanism.
|
|
28
|
+
|
|
29
|
+
The fallback negotiates the installed Paseo surface at runtime with `paseo --version` and help output. It does not assume `--quiet`, JSON output or daemon JSON support. This specifically avoids the failure mode where an older Paseo such as 0.3.1 rejects a flag expected by a newer CLI.
|
|
30
|
+
|
|
31
|
+
Recoverable daemon states such as stale PID, unreachable daemon or connection refused are stopped/restarted and verified before lead/worker launch.
|
|
32
|
+
|
|
33
|
+
## Fresh lead sessions
|
|
34
|
+
|
|
35
|
+
`aeh start` now creates a fresh lead by default:
|
|
36
|
+
|
|
37
|
+
```bash
|
|
38
|
+
aeh start
|
|
39
|
+
```
|
|
40
|
+
|
|
41
|
+
Explicit reuse is:
|
|
42
|
+
|
|
43
|
+
```bash
|
|
44
|
+
aeh start --resume
|
|
45
|
+
```
|
|
46
|
+
|
|
47
|
+
This prevents a new invocation from inheriting stale conversational assumptions, obsolete bootstrap instructions or unnecessary historical context. Durable Git/TaskContract/seal/run/audit/delivery state remains reusable even when the conversation is fresh.
|
|
48
|
+
|
|
49
|
+
## Context pressure and automatic lead rotation
|
|
50
|
+
|
|
51
|
+
Default policy:
|
|
52
|
+
|
|
53
|
+
```text
|
|
54
|
+
<70% normal
|
|
55
|
+
70-80% PRESSURE: stop exploratory work and increase delegation
|
|
56
|
+
>=80% HANDOFF_REQUIRED: write handoff and rotate to a fresh lead
|
|
57
|
+
>=90% HARD_HANDOFF: no further engineering work in the old lead
|
|
58
|
+
```
|
|
59
|
+
|
|
60
|
+
The public check is:
|
|
61
|
+
|
|
62
|
+
```bash
|
|
63
|
+
aeh context guard --agent "$PASEO_AGENT_ID"
|
|
64
|
+
```
|
|
65
|
+
|
|
66
|
+
When Paseo exposes a usable context ratio and the guard runs inside a managed Paseo lead, AEH writes:
|
|
67
|
+
|
|
68
|
+
```text
|
|
69
|
+
.harness/paseo/handoffs/lead-<timestamp>.json
|
|
70
|
+
```
|
|
71
|
+
|
|
72
|
+
and automatically creates a fresh lead. The artifact records prior/new lead IDs, context usage, branch, latest run/audit/delivery references, and an optional semantic brief. The fresh lead is bootstrapped from this deterministic artifact and its referenced sealed state. Normal compaction-and-continue is not the preferred path once the handoff threshold is crossed.
|
|
73
|
+
|
|
74
|
+
If the installed Paseo version does not expose a stable context ratio, AEH returns `UNKNOWN`; it does not fabricate token usage. Fresh-on-start and delegation-first behavior still reduce accumulation.
|
|
75
|
+
|
|
76
|
+
## OpenSpec as SPEC authoring backend
|
|
77
|
+
|
|
78
|
+
For non-QUICK changes, the lead no longer writes proposal/spec/design/tasks itself. `spec-manager` owns SPEC authoring with OpenSpec.
|
|
79
|
+
|
|
80
|
+
```text
|
|
81
|
+
CHANGE
|
|
82
|
+
|
|
|
83
|
+
v
|
|
84
|
+
SPEC
|
|
85
|
+
|
|
|
86
|
+
v
|
|
87
|
+
spec-manager
|
|
88
|
+
|
|
|
89
|
+
|-- aeh spec prepare
|
|
90
|
+
|-- openspec status / instructions
|
|
91
|
+
|-- proposal/specs/design/tasks
|
|
92
|
+
|-- openspec validate --strict
|
|
93
|
+
`-- aeh spec compile
|
|
94
|
+
|
|
|
95
|
+
v
|
|
96
|
+
AEH proposal/spec/design/tasks/Gherkin
|
|
97
|
+
+ TaskContract
|
|
98
|
+
+ stable requirement IDs
|
|
99
|
+
+ validator traceability
|
|
100
|
+
|
|
|
101
|
+
v
|
|
102
|
+
seal
|
|
103
|
+
```
|
|
104
|
+
|
|
105
|
+
Commands:
|
|
106
|
+
|
|
107
|
+
```bash
|
|
108
|
+
aeh spec prepare READABILITY-001 --title "Improve readability"
|
|
109
|
+
aeh spec compile READABILITY-001 --title "Improve readability"
|
|
110
|
+
aeh sdd validate READABILITY-001
|
|
111
|
+
aeh seal READABILITY-001
|
|
112
|
+
aeh run READABILITY-001
|
|
113
|
+
```
|
|
114
|
+
|
|
115
|
+
OpenSpec is an **authoring source before freeze**. The compiled native AEH SDD/TaskContract and seal remain normative during implementation. OpenSpec apply commands do not replace AEH's worker, validation, review-convergence or delivery lifecycle.
|
|
116
|
+
|
|
117
|
+
The compiler records `authoring.provider=openspec`, change name and SHA-256 provenance. It maps OpenSpec `### Requirement:` / `#### Scenario:` content into stable AEH requirement IDs and Gherkin acceptance. Refactor/tooling changes without a delta spec receive an explicit behavior-preservation requirement.
|
|
118
|
+
|
|
119
|
+
A requirement validator may not be invented. AEH uses configured deterministic validators first, otherwise derives a project test/typecheck/build command from known project metadata. If it cannot establish executable requirement evidence, SPEC compilation fails instead of emitting a phantom validator.
|
|
120
|
+
|
|
121
|
+
## Built-in orchestration preset
|
|
122
|
+
|
|
123
|
+
New projects extend:
|
|
124
|
+
|
|
125
|
+
```jsonc
|
|
126
|
+
{
|
|
127
|
+
"extends": ["aeh:orchestration"]
|
|
128
|
+
}
|
|
129
|
+
```
|
|
130
|
+
|
|
131
|
+
This extends `aeh:default`, makes lead/planner delegation-first, and adds:
|
|
132
|
+
|
|
133
|
+
- `environment-manager`;
|
|
134
|
+
- `spec-manager`.
|
|
135
|
+
|
|
136
|
+
OpenSpec is included in the declarative toolchain and automatically selected when `sdd.authoring.provider: openspec` is active.
|
|
137
|
+
|
|
138
|
+
## Trust boundary
|
|
139
|
+
|
|
140
|
+
The authority order remains unchanged:
|
|
141
|
+
|
|
142
|
+
1. user intent / true product decisions;
|
|
143
|
+
2. sealed compiled AEH SDD + TaskContract;
|
|
144
|
+
3. deterministic validators/evidence;
|
|
145
|
+
4. quality/review gates;
|
|
146
|
+
5. LLM summaries and advisory tools.
|
|
147
|
+
|
|
148
|
+
Paseo skills improve communication and lifecycle control; they do not become gate authority. OpenSpec improves specification authoring; it does not become mutable execution truth after the AEH seal.
|