feat(autonomy): prerequisite-resolution slice for the setup skill (#2725)

plugins/autonomy/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://json.schemastore.org/claude-code-plugin-manifest.json",
   "name": "autonomy",
-  "version": "0.21.0",
+  "version": "0.22.0",
   "description": "Governed autonomous agent operation: role-topology, binding-seam, wiring-vs-advisor, telemetry, return-accounting, trigger-dispatch, per-work-class guardrail-matrix, standing-routine-catalog, and design-only runner-charter contracts for climbing the AI-adoption ladder, plus a guided-setup skill that discovers an adopting org's state, writes its schema-versioned binding, wires standards-pinned OTLP emission with a zero-cost file-artifact default, wires human-attested return capture at the task boundary, wires signal adapters with one governed dispatch entrypoint, binds the five-class guardrail matrix to an org's isolation substrates with an in-boundary live-validation probe before recording each fail-closed binding, and stands up standing-routine-catalog classes as scheduled temporal signal adapters behind the one governed queue with free scheduling defaults wired as reviewable changes and each routine's work-class mapping homed on the security surface.",
   "author": {
     "name": "Melodic Software",

plugins/autonomy/CHANGELOG.md

@@ -3,6 +3,19 @@
 All notable changes to the `autonomy` plugin are documented here. Format follows
 [Keep a Changelog](https://keepachangelog.com/en/1.1.0/); this plugin uses semantic versioning.
 
+## [0.22.0]
+
+### Added
+
+- **Prerequisite-resolution setup slice** (#2725). Extends `/autonomy:setup` with
+  `check` (per-identity verdicts + provenance across declared surfaces via
+  `scripts/check-prerequisite-resolution.mjs`; engine health-check liveness) and
+  `apply` (detect-diff-reconcile, prose-context proposals into non-security keys,
+  human ratify → additive `prerequisite_resolution` section with `surface_refs` and
+  no `surfaces` map; narrowing-only enablement; org-rung interviewed never
+  auto-written; security binding prepared never written). Spoke:
+  `context/prerequisite-resolution-slice.md`. Co-located slice test + evals 30–31.
+
 ## [0.21.0]
 
 ### Added

plugins/autonomy/README.md

@@ -51,7 +51,8 @@ state and records that binding.
   prober — with `deferred-class` marking `join:` rows that have no identities yet. Per-identity
   facts live in each `v1` leaf; `generated/identity-prerequisites.json` is the drift-gated
   emission derived from those leaves (generator under `skills/setup/scripts/`);
-  `skills/setup/scripts/resolve-prerequisites.mjs` is the deterministic per-surface resolver.
+  `skills/setup/scripts/resolve-prerequisites.mjs` is the deterministic per-surface resolver;
+  the setup skill's prerequisite-resolution slice (`check` / `apply`) consumes it.
 - **Runner design pack** (`reference/runner.md`): the architect-ready design contract for the
   autonomous-drain runner — the composition spine and its eight seams, the lifecycle state
   model, the two-family stop-criteria taxonomy with terminal-handoff escalation and

plugins/autonomy/skills/setup/SKILL.md

@@ -430,6 +430,35 @@ resolution order.
    `signal.routine`, an unresolvable surface, an identity↔surface mismatch, a raw link outside the
    ratified prefix, a `producer_identity` mismatch, or an unclassified class is a finding.
 
+## Prerequisite-resolution slice
+
+Extends this skill per its own extension model for
+[routine prerequisite resolution](${CLAUDE_PLUGIN_ROOT}/reference/prerequisite-resolution.md).
+Detail lives in
+[`context/prerequisite-resolution-slice.md`](context/prerequisite-resolution-slice.md).
+
+**Liveness.** The slice `check` is an engine health-check surface: it invokes
+[`scripts/resolve-prerequisites.mjs`](scripts/resolve-prerequisites.mjs) end-to-end and
+fails loud on internal failure — never a verdict-shaped fallback.
+
+1. **`check`** — for each scheduling surface already recorded under `triggers` / `routines`
+   (this slice declares no `surfaces` map of its own), report per-identity verdicts with
+   provenance via
+   [`scripts/check-prerequisite-resolution.mjs`](scripts/check-prerequisite-resolution.mjs).
+   A bare repo yields `unsupported` / `unknown` for every identity, never an error.
+2. **`apply`** — detect-diff-reconcile against existing `prerequisite_resolution`
+   declarations; a declaration contradicting a ran-negative probe is a finding (identity
+   stays negative while the finding is open — ADR 0011 Decision 2). The prose-context pass
+   reads `CLAUDE.md` / `AGENTS.md` (session-reachable only by reference) / `README` to
+   *propose* declarations into **non-security keys only**; the human ratifies; the slice
+   writes the additive section (`surface_refs` + `declarations`, no `surfaces` map).
+   Narrowing-only enablement: enable in `routines.enabled` only when the verdict clears;
+   negative/`unknown` routes to the advisory path. Org-rung entitlements are interviewed,
+   never auto-written. Non-interactive contexts skip ask-and-persist and report assumptions.
+   Security-binding changes are **prepared**, never written.
+   Wrapper:
+   [`scripts/apply-prerequisite-resolution.mjs`](scripts/apply-prerequisite-resolution.mjs).
+
 ## Runner note
 
 The [runner design pack](${CLAUDE_PLUGIN_ROOT}/reference/runner.md) is bindable-when-born:
@@ -454,7 +483,8 @@ surface recorded in two `surfaces` maps resolving as ambiguous — are catalogue
 ## What this skill does NOT do
 
 - Wire capability slices that have not shipped yet — each lands with its own work package and
-  extends this skill (the runner charter execution pack is the next such slice).
+  extends this skill (the runner charter execution pack remains the next such slice after
+  prerequisite-resolution).
 - Estimate, impute, or backfill the two human-attested return fields — ever.
 - Mutate platform settings, user settings, or `pluginConfigs`.
 - Assume the shape of any particular org or fleet — a run against an unknown repo asks or

plugins/autonomy/skills/setup/context/prerequisite-resolution-slice.md

@@ -0,0 +1,81 @@
+# Prerequisite-resolution slice
+
+Extends `/autonomy:setup` per the skill's own extension model. Owns the
+[routine prerequisite resolution](${CLAUDE_PLUGIN_ROOT}/reference/prerequisite-resolution.md)
+question at setup time: which `v1` identities can run against this repository on each
+declared scheduling surface, and why.
+
+## Liveness
+
+This slice's `check` is an **engine health-check** surface under
+[`liveness-assertion`](../../../../../docs/conventions/liveness-assertion/README.md): it
+invokes the deterministic resolver end-to-end and fails loud on internal failure. It never
+reports "healthy" from configuration alone, and never invents a verdict-shaped fallback.
+
+## `check` (read-only)
+
+1. Resolve scheduling-surface ids from the existing binding (`triggers.surfaces` and
+   `routines.surfaces` — merged; the slice never declares its own `surfaces` map).
+2. For each surface, run
+   [`scripts/resolve-prerequisites.mjs`](../scripts/resolve-prerequisites.mjs) against the
+   project root.
+3. Report per-identity verdicts (`supported` / `conditional` / `unsupported` / `unknown`)
+   with per-signal provenance and any findings (declaration↔probe contradictions).
+4. On a bare repo (no binding, no tracker, no CI), every identity reports
+   `unsupported` or `unknown` — never an error.
+
+Wrapper:
+[`scripts/check-prerequisite-resolution.mjs`](../scripts/check-prerequisite-resolution.mjs).
+
+## `apply` (interactive propose → ratify)
+
+1. **Detect-diff-reconcile.** Run the same resolution as `check`. An existing
+   `prerequisite_resolution` declaration is authoritative input: divergence from probe
+   results is a **finding**, never a silent overwrite. A ran-negative probe caps a positive
+   declaration (ADR 0011 Decision 2) — the identity stays `unsupported` while the finding
+   is open.
+2. **Prose-context pass (proposal only).** Read host instruction files (`CLAUDE.md`),
+   secondary agent-instruction files (`AGENTS.md` — reaches a session only through a
+   reference), and `README` for *proposed* declarations into **non-security keys only**.
+   The deterministic resolver never parses prose; prose is never runtime authority.
+3. **Human ratifies.** Interactive contexts present proposals one at a time. Non-interactive
+   and forked contexts skip ask-and-persist rungs and report assumptions (topic-docs rule).
+4. **Write additively.** On ratification, write the `prerequisite_resolution` section of
+   `.claude/autonomy/binding.json`:
+   - `schema_version`: `"1.0"`
+   - `surface_refs`: existing scheduling-surface ids (references only — **no `surfaces` map**)
+   - `declarations`: `{ surface, identity?, need?, state, rung }` entries
+5. **Narrowing-only enablement.** An identity may be enabled in `routines.enabled` only when
+   its verdict clears (`supported`, or `conditional` where the named conditions are accepted).
+   `unsupported` / `unknown` route to the advisory path. The slice **prepares** any
+   security-binding change (admission / classification) and **never writes** that surface.
+6. **Org-rung entitlements.** Connector entitlements for `prod` / `product` / `org` / `ext`
+   bind at the Org binding layer. The slice reports which prerequisites await the org rung
+   and stops — it never auto-writes org-rung values into the repo-local binding.
+
+Wrapper (non-interactive propose / optional `--ratify` for tests):
+[`scripts/apply-prerequisite-resolution.mjs`](../scripts/apply-prerequisite-resolution.mjs).
+
+## Binding section shape
+
+```json
+{
+  "prerequisite_resolution": {
+    "schema_version": "1.0",
+    "surface_refs": ["ci-cron"],
+    "declarations": [
+      {
+        "surface": "ci-cron",
+        "identity": "issue-triage-sweep",
+        "need": "tracker",
+        "state": "present",
+        "rung": "repo-local"
+      }
+    ]
+  }
+}
+```
+
+Absent-section tolerance holds. The section MUST NOT carry a `surfaces` map —
+[`check-signal-envelope.mjs`](../scripts/check-signal-envelope.mjs) merges every section's
+`surfaces` map and treats duplicates as ambiguous.

plugins/autonomy/skills/setup/evals/evals.json

@@ -378,6 +378,32 @@
         "Restates that the runner-execution home is unborn and nothing else runner-specific is recorded until a build trigger fires",
         "Cites the runner design pack rather than restating its stop-criteria taxonomy or severity-resolution content"
       ]
+    },
+    {
+      "id": 30,
+      "name": "prerequisite-resolution-check-reports-verdicts-with-provenance",
+      "prompt": "/autonomy:setup check\n\nWhich routine identities can actually run against this repo, and why?",
+      "expected_output": "Runs the prerequisite-resolution slice check: for each scheduling surface already recorded in the autonomy binding (the slice declares no surfaces map of its own), invokes the deterministic resolver and reports per-identity verdicts (supported / conditional / unsupported / unknown) with per-signal provenance and any declaration↔probe findings. Names its liveness taxonomy row (engine health-check) and fails loud on resolver failure. On a bare repo every identity is unsupported or unknown — never an error. Performs no writes.",
+      "files": [],
+      "expectations": [
+        "Reports per-identity verdicts with provenance across declared scheduling surfaces",
+        "States engine health-check liveness taxonomy row / fail-loud posture",
+        "Bare-repo path yields unsupported/unknown for every identity without error",
+        "No file writes in check mode"
+      ]
+    },
+    {
+      "id": 31,
+      "name": "prerequisite-resolution-apply-reconcile-narrowing-no-security-write",
+      "prompt": "/autonomy:setup apply\n\nDetect our routine prerequisites from CLAUDE.md and enable everything that looks ready.",
+      "expected_output": "Runs detect-diff-reconcile: existing prerequisite_resolution declarations are authoritative; a declaration contradicting a ran-negative probe is a finding (identity stays unsupported while the finding is open — never a silent overwrite, and a positive verdict never survives on intent alone). The prose-context pass proposes declarations into non-security keys only from CLAUDE.md/AGENTS.md/README; the human ratifies before any write. The additive binding section uses surface_refs and declarations and declares no surfaces map. Enablement is narrowing-only — unsupported/unknown route to the advisory path. Org-rung entitlements are interviewed and never auto-written. Security-binding changes are prepared, never written. Non-interactive contexts skip ask-and-persist and report assumptions.",
+      "files": [],
+      "expectations": [
+        "Declaration↔probe contradiction surfaces as a reconcile finding; identity remains unsupported",
+        "Prose proposals target non-security keys only; human ratification before write",
+        "prerequisite_resolution section has surface_refs + declarations and no surfaces map",
+        "Never writes the security binding; org-rung entitlements interviewed not auto-written; narrowing-only enablement"
+      ]
     }
   ]
 }