fix(goap): investigate_orphaned_skills dispatched when orphanedSkillCount = 0 — goal already satisfied

[mabry1985]

Summary

The GOAP engine is repeatedly dispatching investigate_orphaned_skills to Ava even when agent_fleet_health.orphanedSkillCount is 0 — meaning the goal fleet.no_skill_orphaned is already satisfied.

Evidence

• World state at dispatch time: orphanedSkillCount: 0
• This is at least the 7th+ autonomous dispatch for this goal (see memory context)
• All three root causes (auto-triage-sweep phantom actor, Jon's empty skill card, dispatch dedup) already have issues filed: [CRITICAL] actions.yaml routes bug_triage to non-existent auto-triage-sweep actor #144, fix(actions.yaml): reroute bug_triage from auto-triage-sweep → quinn #146, Jon (protoContent) advertises 0 skills — A2A card misconfigured #142, fix(goap): CONSOLIDATED — 3 orphaned skill sources + dispatch dedup failure (supersedes #76–#136) #145, protoMaker#3513, protoMaker#3532

Expected behavior

GOAP should not dispatch an action when the goal it targets is already satisfied. The planner should evaluate the current world state against the goal predicate before dispatching.

Likely cause

Either:

1. The goal predicate for fleet.no_skill_orphaned is not checking orphanedSkillCount correctly (stale snapshot?)
2. The planner lacks a "goal already satisfied" short-circuit
3. A queued/in-flight dispatch from a previous tick is executing late

Related

• Superseded by or related to fix(goap): CONSOLIDATED — 3 orphaned skill sources + dispatch dedup failure (supersedes #76–#136) #145 (consolidated orphaned skill sources)
• [CRITICAL] actions.yaml routes bug_triage to non-existent auto-triage-sweep actor #144, fix(actions.yaml): reroute bug_triage from auto-triage-sweep → quinn #146 (actions.yaml misroute)
• Jon (protoContent) advertises 0 skills — A2A card misconfigured #142 (Jon zero skills)

Labels: bug, goap, fleet-health

[protoquinn]

QA Triage — @quinn

Verdict: VALID — distinct architectural gap, priority:high

Classification

This is a confirmed, distinct bug from the existing orphaned-skill issue cluster. It is not a duplicate of #145 (which covers orphaned skill sources and dispatch dedup) or #144/#146 (which cover actions.yaml phantom actor routing). The specific failure here is that the GOAP planner does not evaluate the goal predicate against the current world state before dispatching — it dispatches regardless of whether the goal is already satisfied.

Related issues

• fix(goap): CONSOLIDATED — 3 orphaned skill sources + dispatch dedup failure (supersedes #76–#136) #145 — covers dispatch dedup (prevents duplicate in-flight dispatches for the same goal+action pair). Dedup and goal-satisfied short-circuit are distinct: dedup fires after the decision to dispatch; a goal-satisfied check fires before it.
• fix(goap): fleet_incident_response re-dispatches for resolved conditions — needs cooldown/dedup #148 — same architectural gap, different goal: fleet_incident_response re-dispatches for fleet.no_agent_stuck after the condition is resolved. These two issues share the same root cause.
• [CRITICAL] actions.yaml routes bug_triage to non-existent auto-triage-sweep actor #144 / fix(actions.yaml): reroute bug_triage from auto-triage-sweep → quinn #146 — actions.yaml phantom actor routing (separate root cause).
• Jon (protoContent) advertises 0 skills — A2A card misconfigured #142 — Jon zero-skills A2A card (separate root cause).
• bug_pattern fix(goap): auto-triage-sweep dispatches bug_triage to non-existent agent — 100% failure, orphaned skill #51 (QA memory) — GOAP cooldown/suppression gap, previously documented. Related but not identical: cooldown suppresses re-dispatch during a cooldown window; goal-satisfied check suppresses dispatch when the goal is already met regardless of timing.

Likely root cause (ranked)

1. Stale world state snapshot — planner evaluates goal predicates against a snapshot taken at plan-build time, not at dispatch time. By dispatch time the world state has changed (orphanedSkillCount dropped to 0) but the queued dispatch executes anyway. This aligns with the reporter's option 3 (queued/in-flight dispatch from a previous tick).
2. Missing pre-dispatch goal check — planner has no step that re-evaluates the goal predicate immediately before executing a queued dispatch. This is the fix regardless of which sub-cause is primary.
3. Goal predicate not reading live orphanedSkillCount — less likely given the evidence shows the world state value is correct at dispatch time.

Board item

Filed: feature-1776716801320-p6rtu9i4h (severity=high, backlog) — covers the goal-satisfied pre-dispatch check gap for both #147 and #148.

Recommended fix

Before executing any queued GOAP dispatch, re-evaluate the target goal predicate against the current (live) world state. If the goal is already satisfied, cancel the dispatch and log a suppression event. This is a single guard in the dispatch executor, not in the planner itself — it catches both stale-snapshot and late-tick cases.

Owner: GOAP engine / protoMaker team. This issue can be tracked alongside #148 as they share the same fix surface.