Skip to content

Historical field note

This record preserves prior investigation or planning evidence. It is not a current operating contract; use the numbered book and current runbooks for instructions.

Triage: Dream Finding dreamrepair:828e1ef4a530935a9a7db4b1807202e1

Read-only corroboration for triage task task_bd52602a4ea446f1a9ca1936f9d160c2 (title: "Triage dream finding dreamrepair:828e1ef4: corroborate the single evidence record"), a cooperative child of the parent audit task_bcbebedf59404975a16fb17f6f9f6032 ("Investigate low-confidence dream finding: skill"). This document records ground truth only. No src/, tests/, skills/, or deploy/ file was changed. All facts below were read from the live MAC hub read API (GET /health, GET /memory, GET /tasks/{id}).

Summary Verdict

NOT ACTIONABLE — evidence gap. The finding's sole supporting record, mem_9c807de962da4ecda4eac62670006672, is a mislabeled [success] plan_decomposed record, not a failure. The 'skill' label is a generic regex placeholder (\bskill[s]?\b), not a concrete named skill or defect. The evidence is self-referential across a chain of prior identical investigation tasks: each "Investigate low-confidence dream finding: skill" task, when decomposed, emits a success record that the nap/dream cycle re-classifies as a new low-confidence failure_pattern, spawning the next identical task. There is no reproducible defect for a code or skill change to fix.

Corroboration Method (hub read API)

  • GET $MAC_HUB_URL/health → {"status":"ok"}.
  • GET /memory?task_id=task_5568db6185174de5ab295031da3246d2&limit=100 returned exactly 3 records for the candidate task; the target record is present.
  • GET /tasks/task_5568db6185174de5ab295031da3246d2 and GET /tasks/task_70007fd6f9a64bc8bb434873f3c3dc9c supplied the metadata.dream_repair classification blocks used below.

The Single Supporting Evidence Record

mem_9c807de962da4ecda4eac62670006672 - task_id: task_5568db6185174de5ab295031da3246d2 - record_type: deployment_learning:mac; created_by: mac-task-executor - content (mac.deployment_learning.v1): evidence_type = "plan_decomposed", outcome = "success", error_signature = "", signals = {returncode: 0, tests: null, ...}, task_title = "Investigate low-confidence dream finding: skill".

The derived dream artifact mem_ba0c35995e20434f9097d313a5d77b97 (dream:failure_pattern, mac.dream.v1, confidence_score = 0.35, evidence_count = 1) records the observation verbatim as: "[success] Investigate low-confidence dream finding: skill (plan_decomposed)".

Answers to the Three Triage Questions

  1. Genuine failure or a mislabeled [success] plan_decomposed record? Mislabeled. The record's own outcome is success and its evidence_type is plan_decomposed (a task-decomposition bookkeeping event, returncode = 0, empty error_signature). The failure_pattern kind comes from the dream classifier grouping, not from any observed failure.

  2. Is 'skill' a concrete named defect or a generic regex placeholder? Generic placeholder. In src/mac/dream_cycle_classifier.py the first _SKILL_PATTERNS entry (r"\bskill[s]?\b", "skill") maps any occurrence of the bare word "skill"/"skills" to the canonical area_name = "skill". The candidate summary literally contains the token ("...dream finding: skill..."), so the label is a self-fulfilling word-boundary match, not a named skill. affected_tools/providers/repo_areas are all empty; the two live SKILL.md files (skills/mac-agent-terminal-timeout, skills/setup-mac-fleet) are healthy.

  3. Is the evidence self-referential across a chain of prior identical investigation tasks? Yes. Traced links (each finding's sole evidence is the prior identical task's plan_decomposed success):

  4. dreamrepair:828e1ef4a530935a9a7db4b1807202e1 (this finding) ← candidate task_5568db6185174de5ab295031da3246d2 ← evidence mem_9c807de962da4ecda4eac62670006672 (plan_decomposed/success).
  5. task_5568db...'s own metadata.dream_repair = dreamrepair:6b7c8892c8628c6d28a51438528b904d, candidate task_70007fd6f9a64bc8bb434873f3c3dc9c, evidence mem_5c221c62251c4d0688d153276b766c7b (plan_decomposed/success).
  6. task_70007fd6...'s own metadata.dream_repair = dreamrepair:cf8fa7ce5d85c7e0922d419443055a43, candidate task_8764c80cdff34d20867478306b7387a0, evidence mem_22ba025a49a0453483d1ca0882181893. All share the identical title "Investigate low-confidence dream finding: skill" and the identical summary shape "...Supported by 1 memory record(s): [success] ... (plan_decomposed)". This is a feedback loop: the pattern re-derives itself from its own prior decomposition.

Recorded Evidence Gap

Support is effectively zero net-new: a single low-confidence (0.35, evidence_count = 1), self-referential record that is a plan_decomposed success, labeled by a generic \bskill[s]?\b regex with no named skill/tool/provider/repo-area and no reproducible failure signature (empty error_signature, returncode = 0). Confidence 0.35 reflects support < 2, not a diagnosed fault.

Disposition Handoff

  • Decision: NOT ACTIONABLE. Recommend closing dreamrepair:828e1ef4a530935a9a7db4b1807202e1 and treating the identical-title chain as a classifier feedback loop rather than fresh failure evidence.
  • Reopen criteria: a named skill or tool acquiring a reproducible failure signature (real error/stack/failing test) with at least two independent, non-self-referential evidence records.
  • No skill/tool changes were made, per task scope.