Execution Log: emoji reactions acknowledging that the-loop is processing an entity

Append-only log of progress for the user's visibility. Checked in alongside the spec at docs/specs/issue-84/.

Phase transitions

Phase	Entered	Reviewed/approved by	Notes
requirements-definition	2026-07-24		Issue #84 body: 👀 started / ✅ completed / ⁉️ error, CLI-config-driven, no-op where unsupported. GitHub-palette constraint surfaced (✅/⁉️ unavailable → closest defaults, configurable).
design	2026-07-24		Dispatcher-level seam (covers webhook + poll); reactions.py (config, target resolution, gh api reactor); default off (first daemon write surface).
tasks-breakdown	2026-07-24		10-task DAG
implementation	2026-07-24		Implemented on claude/github-issue-84-me696x
needs-review	2026-07-24		PR opened; awaiting human review (tier-3, human-approves-pr)
complete

Progress entries

2026-07-24 — spec drafted, implementation started

• Phase: requirements → design → tasks → implementation
• Did: Read the trigger path end-to-end (router → dispatcher → poller). Established that both ingress paths share the dispatcher, making it the one seam where "started/completed/error" is truthfully observable; confirmed the poll path's comment ids are GraphQL node ids while webhook ids are numeric (drives the REST/GraphQL split). Drafted the 3-phase spec.
• Next: implement T1–T10.
• Blockers: none. Human approval of the spec + code happens at the PR (tier-3, human-approves-pr).

2026-07-24 — implemented (T1–T10)

• Phase: implementation → needs-review
• Did:
  • reactions.py: ReactionConfig, target_from_event, GitHubReactor (REST + GraphQL via gh api, warn-once missing gh, never raises).
  • webhook/dispatcher.py: RoutingConfig.reactions; injectable reactor surviving reload; _worker reacts started/completed/error; _dispatch_one/_spawn_for/_spawn_tmux/_respawn_tmux return bool.
  • eventlog.py: registered reaction.added / reaction.failed.
  • Config: routing.reactions in cli-config.schema.json (enum-guarded) + dogfood block in cli-config.yaml (enabled: true).
  • Tests: test_reactions.py unit suite + dispatcher integration scenarios (success/failure/spawn/no-reaction paths) with a recording fake reactor.
  • Docs: cli/README.md reactions reference; webhook-triggers.md behaviour and history row (the observability reference defers to EVENT_TYPES, which a unit test keeps in sync — no edit needed).
• Checkpoint/tests: make check green — ruff, markdownlint, ruff format, pyright 0 errors, validate_config.py VALID, pytest all passed.
• Next: open PR with the reviewer briefing; request review.
• Blockers: none.

2026-07-24 — owner decision on PR #85: default-on

• Decision: the owner answered the briefing's open question — reactions default to enabled: true (was drafted default-off as the conservative posture for the daemon's first GitHub write). Paper trail: the owner's comment on PR #85 ("default should be on").
• Did: flipped the default in ReactionConfig (+ from_mapping), the config schema, both cli-config.yamls (repo + template), README, capability doc and this spec (AC2.3, security section, design decision). Added a conftest autouse fixture stubbing the dispatcher's default reactor so the suite stays hermetic now that a bare RoutingConfig() enables reactions; updated the default-asserting tests.
• Checkpoint/tests: make check green again.

Review cycles

Cycle	Type (self/critic)	Reviewer	Outcome	Link

Final validation evidence

Recorded on the PR: local gate output (ruff/markdownlint/pyright/validate/ pytest) and the unit/integration scenarios demonstrating each acceptance criterion (started/completed/error targets, config gating, no-op paths).