feat(skills): harness-compatibility sweep in orchestration-prompt

packages/tapps-mcp/src/tapps_mcp/pipeline/platform_skill_orchestration.py

@@ -169,6 +169,15 @@
 - **Self-healing preconditions.** Anything the loop needs (a runtime up, a
   scorer/tool built, a branch, auth reachable) is a **Sub-goal 0** the loop
   *establishes itself* — never a "set this up first" note the user must action.
+- **Harness-compatibility sweep.** The runner session carries the *project's own*
+  harness: PreToolUse/PostToolUse hooks that gate tool calls (issue-tracker write
+  sentinels, prod guards) and MCP-server standing instructions that nudge per-edit
+  behavior (quality checks after every file edit, doc lookups). Enumerate the gates
+  and nudges the loop's tool calls will actually hit; bake each required
+  unlock/refresh step into Sub-goal 0 or the relevant loop step, and in Guardrails
+  explicitly **adopt or override** each standing nudge (e.g. "quality pipeline runs
+  at the epic gate, not per edit — this overrides the per-edit nudge"). A prompt
+  that fights its own project's hooks burns its budget on diagnose loops.
 - **Deploy-freshness + smoke/health gate** (any prompt that runs against a live or
   deployed target, not source): in Sub-goal 0, self-healing — (1) **merged ≠ live**:
   if the target is a baked image, compare latest merged commit to the build time and
@@ -199,6 +208,9 @@
   Workflow `budget` to a token ceiling (≈ the autonomy cost gate) so it self-aborts.
 - **Memory** — recall at the start, record the outcome (incl. failures) at each
   checkpoint, so learning survives the session.
+- **Harness compatibility** — every tool call the loop makes that is gated by a
+  project hook has its unlock/refresh step in the prompt, and every MCP standing
+  nudge is explicitly adopted or overridden (method §6).
 
 ## Autonomy contract (every emitted prompt carries this)
 
@@ -248,7 +260,9 @@
    return + per-agent contract; a memory recall+record step; an **Autonomy
    contract**; a **bounded diagnose-don't-repeat** path; a **context-hygiene** line;
    and the **Engineering discipline** line. For a live/deployed target, confirm
-   Sub-goal 0 has the deploy-freshness + smoke/health gate. Run the **cold-start
+   Sub-goal 0 has the deploy-freshness + smoke/health gate. Confirm **harness
+   compatibility**: every hook-gated tool call has its unlock/refresh step and every
+   MCP standing nudge is adopted-or-overridden. Run the **cold-start
    test**: a fresh session with nothing loaded can run it. Fix anything weak before
    saving.
 6. Tell the user exactly how to run it — the `/goal` line, the `/loop` cadence, the
@@ -294,6 +308,7 @@
 0. **Establish preconditions (self-healing — the loop sets these up, NOT the user).** <runtime up, scorer/tool built, auth reachable, branch ready>
    - **Deploy freshness (live/deployed target only):** merged ≠ live. If baked image, compare latest merged commit to build time; rebuild/redeploy (preserve overlays) if `main` is newer. Stale image = required-fail cap.
    - **Smoke + health gate (after any deploy, before the real run):** `/health` is `ok|degraded` and one cheap end-to-end call succeeds.
+   - **Harness compatibility:** <PreToolUse gates + MCP standing nudges the loop's tool calls will hit → bake unlock/refresh steps here; adopt-or-override each nudge in Guardrails>
    - proof: <preconditions verified; for live targets — image no older than latest merged commit + a 200/non-error smoke pasted>
 1. <narrow, verifiable> — proof: <ground-truth artifact>
 2. <…>
@@ -323,6 +338,7 @@
 - Context hygiene — targeted grep over full re-Read.
 - Scope: repos in play = <list>; reads fleet-wide, writes via owner.
 - Memory: recall at start; record outcome (incl. failures) at each checkpoint.
+- Harness compatibility: <gated tool calls → unlock/refresh steps; MCP standing nudges → adopted or overridden>.
 - Discipline: root-cause not workarounds; no green-by-suppression; right-sized; durable; match conventions; no scope creep.
 
 ## Autonomy

packages/tapps-mcp/tests/unit/test_orchestration_prompt_skill.py

@@ -61,6 +61,17 @@ def test_template_has_verifier_and_tier_columns(self, tmp_path):
         assert "model tier" in tpl
         assert "verifier subagent" in tpl
 
+    def test_body_and_template_carry_harness_compatibility(self, tmp_path):
+        generate_skills(tmp_path, "claude")
+        d = _skill_dir(tmp_path)
+        content = (d / "SKILL.md").read_text().lower()
+        # method §6: the emitted prompt must survive the project's own hooks
+        # (PreToolUse gates) and MCP standing nudges — adopt or override each.
+        assert "harness-compatibility sweep" in content
+        assert "adopt or override" in content or "adopted or overridden" in content
+        tpl = (d / "assets" / "prompt-template.md").read_text().lower()
+        assert "harness compatibility" in tpl
+
     def test_cursor_host_also_gets_skill(self, tmp_path):
         generate_skills(tmp_path, "cursor")
         assert (_skill_dir(tmp_path, "cursor") / "SKILL.md").exists()