workflow-fix: Per-tick push cap in watcher triage observer
Highlight text on any card (body, plan, or an event) to anchor a comment, or leave a whole-task comment here. Mention @claude to summon a reply.
No comments yet.
Overview / Motivation
Auto-filed by the /daily Step-C parked-candidate sweep (2026-07-08) from a candidate parked on task #967 by a recursion-guarded workflow-fix session.
Goal
Add a per-tick PUSH cap in _triage_observer_emit mirroring the existing per-tick marker cap: push the first K actions, then one summary push ('+N more, see sidecar'); overflow stays sidecar-recorded.
Workflow gap
- Bug observed: _triage_observer_emit sends one Telegram push per flagged action with no per-tick cap (only the epm:progress marker channel is capped) — any backlog (the original ~70-candidate first-tick flush, or a multi-hour watcher outage followed by a burst of matured dispatch records) floods pushes in one tick.
- Why it is a workflow gap: The observer is observe/alert-only by contract; an unbounded push channel turns a backlog into a notification storm that trains the user to ignore the channel.
- Confidence (emitter): medium
- Sweep verification (2026-07-08): Verified 2026-07-08: _triage_observer_emit (autonomous_session_watch.py line 4074) calls _telegram_push per action under
if a['push']with no counter; only markers_posted is capped. Note the motivating one-time first-tick 48h-lookback flush has long since fired — residual value is bounding post-outage bursts, so this is real but low-urgency; the planner may right-size K accordingly.
Proposed change (candidate diff sketch — refine in planning)
In _triage_observer_emit: thread a pushes_sent counter + a _TRIAGE_OBSERVER_PUSH_CAP (env-overridable) mirroring the marker cap; on overflow send one digest push naming the overflow count and the sidecar path.
Scope / surfaces
- Primary target:
scripts/autonomous_session_watch.py - Grep the workflow surface for the pattern before editing (
grep -rln '<pattern>' .claude/ CLAUDE.md scripts/) and update every hit; list them in the plan.
Constraints / invariants
- Workflow-surface only — never experiment code,
configs/, ortasks/. scripts/workflow_lint.py --check-askspasses; ruff on touched files passes; ifworkflow.yamlorCLAUDE.mdchange, they stay consistent with the rule file.- This session runs under
EPM_WORKFLOW_FIX_SESSION=1and carries aworkflow_fix_target:Provenance line — it MUST NOT auto-route any of its own subagents' workflow-fix candidates (recursion guard, workflow-fix-on-bug.md).
Provenance
- workflow_fix_target: scripts/autonomous_session_watch.py
- origin: parked candidate on task #967 at 2026-07-04T08:40:38Z
Verbatim parked note:
source: prose-followup (implementer round 1). target_file: scripts/autonomous_session_watch.py — add a per-tick PUSH cap in _triage_observer_emit mirroring the existing marker cap, bounding the one-time first-tick backlog flush (~70 pushes from the live 48h lookback). routed: parked — running under workflow_fix_target Provenance (recursion guard, .claude/rules/workflow-fix-on-bug.md § Recursion guard); logged for the parent orchestrator / next PM pass.