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.

Disposition: low-confidence dream finding slack (dreamrepair:cc1dedb0d3036d289aafc1e42b4a22aa) — not actionable

Finding: a low-confidence dream-repair failure_pattern, scope project, project mac, provider label slack, fingerprint dreamrepair:cc1dedb0d3036d289aafc1e42b4a22aa, confidence low (overall_confidence_score = 0.35), backed by a single evidence record. Evidence source: this disposition consumes the evidence-review summary docs/archive/field-notes/investigation-dreamrepair-cc1dedb0-slack.md, which characterized the sole supporting record mem_843510f5c2344eab881036d529b392a5 (deployment_learning:mac) from originating task task_2a25617199eb43dd8cf95de6c40ef0a1 ("Convert fleet config to OpenClaw terminology and constrain Slack/persona runtime to openclaw|none"). Dispositioned by: fleet worker (disposition only; no production code, test, skill, config, or deploy edits).

Disposition: CLOSE — NOT ACTIONABLE (evidence-provenance artifact)

Close the finding as not actionable as a defect of the mac chat-gateway, persona, or Slack runtime surface. The finding is the deterministic output of the dream/nap consolidation heuristics acting on a single preliminary, executor-authored self-snapshot that was superseded within the same task run. No concrete, reproducible defect exists for a config, allow-list, terminology, or code change to fix, so the correct handling is to record this disposition as investigation evidence and make no change to skills, tools, config, or source.

Acceptance-criteria evaluation

  • (1) Is the failure_pattern real and recurring, or a single low-confidence artifact? A single low-confidence artifact. The record is explicitly "Supported by 1 memory record(s)"; the 0.35 score is the classifier's deterministic single-record structural floor for support < 2 (see the low tier documented at src/mac/dream_cycle_classifier.py:15 and src/mac/dream_cycle_classifier.py:87), not a corroborated fault. The later dream:failure_pattern re-emissions are the same nap window re-scoring the same lone record, not independent occurrences.

  • (2) Is the Slack provider association signal or noise, given no skills/tools/repo-areas were flagged? Noise (incidental). The slack provider label is echoed verbatim from the originating task's title/description wording ("constrain Slack/persona runtime"), not derived from any Slack transport, chat-gateway, or persona-binding failure. With no skill, tool, or repo_area co-flagged, there is nothing that localizes a Slack-surface defect.

  • (3) Smallest repair vs. follow-up plan. Neither is warranted as a product fix. The change the originating task was about landed and is intact on main: the persona/Slack runtime selector allow-list {openclaw, none} is present and enforced (src/mac/fleet_setup.py:208, "gateway_impl must be openclaw or none"), reading the OpenClaw defaults block with a legacy hermes: read-only fallback (src/mac/fleet_setup.py:194). There is no failing suite, assertion, trace, or reproducer tied to this finding to repair.

Why the sole evidence is not a defect

Per the evidence-review, mem_843510f5c2344eab881036d529b392a5 is a mid-run executor self-snapshot: outcome=failure with returncode=0, tests=fail, pushed=false — the shape the executor records before the contract suite is finalized and before the guarded push, not the shape of a terminal failure. Its error_signature is the run's own prose change summary (no offending module, no failing test id, no reproduction). The originating task then went needs_review → reviewing → approved → published to git://main → completed, and a second deployment_learning:mac record for the same task_id (mem_627afb5e4e2c4f09b495eea8984a7b10, authored by hub-review-workflow) records the authoritative outcome=approved_published. The finding rests on the earlier, superseded snapshot only.

Evidence gap — what would raise confidence

This disposition is a close, not a permanent verdict. State the gap plainly: one supporting record, confidence 0.35, a preliminary self-snapshot rather than a terminal outcome, and a generic slack label with no named skill, tool, or repo_area. Reopen only if independent, reproducible evidence appears that the Slack/chat-gateway/persona surface actually misbehaves — concretely, any of:

  • A failing contract test or reproducer pinning a concrete defect in the Slack transport, chat-gateway runtime, or persona binding, reproducible via scripts/run-contract-tests.sh (e.g. against src/mac/_hermes/gateway/platforms/slack.py or src/mac/fleet_setup.py).
  • A concrete defect location replacing the incidental slack provider label — an actual failing module/tool/repo_area, not task-title wording.
  • Independent corroboration raising evidence_count to >= 2 from records that are not the same run's preliminary self-snapshot, so confidence can reach medium/high on its own merits.

Durable, non-Slack takeaway (out of scope here)

The real generic issue is an evidence-provenance gap in the dream pipeline, not a Slack defect: the classifier keyed on an executor preliminary deployment_learning snapshot (outcome=failure, pushed=false) and did not reconcile it against the terminal approved_published record for the same task_id (emitted at src/mac/services.py:21547). A durable improvement belongs to the classifier/consolidator — prefer the latest terminal outcome per task_id, or suppress preliminary failure snapshots once a later approved_published record exists for the same task. This is recorded as a follow-up consideration only and is explicitly not implemented here, per the directive not to change skills, tools, or config.

Verification

Independently re-confirmed in the task-owned worktree with the bootstrapped .venv (all sources read, not modified): the persona/Slack runtime allow-list and OpenClaw fallback are present and enforced in src/mac/fleet_setup.py, the 0.35 single-record confidence floor and area vocabulary are documented in src/mac/dream_cycle_classifier.py, and the terminal approved_published outcome is emitted by src/mac/services.py. No src/, tests/, skills/, .mac/, or deploy/ file was modified by this disposition; the only artifact is this field note.