pulse

Session Status

Provide a compact summary of the current session state. Designed for channel responses (Telegram, Discord) where brevity matters.

Plan

1. Read .claude-code-hermit/sessions/SHELL.md
2. If the file does not exist: respond "No active session. Run /claude-code-hermit:session to start one." 2b. If Status is idle (session between tasks), format as:

   Session (idle) | started YYYY-MM-DD | N tasks completed
   Last: [latest Session Summary entry]
   Ready for work. Tell me what's next, or run /claude-code-hermit:session-start
   Cost: $X.XX (cumulative)
   

   Return this output and stop — do not proceed to step 3.
3. Parse the following fields from SHELL.md:
   • ID from **ID:** line
   • Status from **Status:** line
   • Tags from **Tags:** line (if present and non-empty)
   • Budget from **Budget:** line (if present and non-empty)
   • Task — first non-comment, non-empty line after ## Task
   • Task progress — call TaskList and count by status. Total = all tasks, completed = completed tasks.
   • Current step — first task with status in_progress
   • Blockers — content under ## Blockers (if any non-comment content)
   • Cost — content under ## Cost (if any non-comment content)
4. Format as a compact output (under 10 lines):

Session S-NNN | in_progress | [tags if present]
Working on: one-line summary
Progress: X/Y tasks | Current: Step N - description
Budget: $spent / $total (percentage%)
Blockers: none (or brief description)
Cost: $X.XX (NNK tokens)

• Omit the Budget line if no budget is set
• Omit tags from the header if none are set
• If the session is blocked, append: "Run /debug to diagnose, or /claude-code-hermit:session to start a new session."

--full flag

When the operator passes --full, append infrastructure health sections after the session block above. This answers "is my hermit healthy?" — not a config dump (that's hermit-settings), not a work summary (that's brief).

5. Read .claude-code-hermit/config.json
6. Glob .claude-code-hermit/proposals/PROP-*.md. For each file, parse the status field from YAML frontmatter (or bullet-point metadata for legacy files). Count per status (proposed, accepted, in_progress, resolved, dismissed, deferred).
7. Read .claude-code-hermit/state/micro-proposals.json. Count entries where status is pending.
8. From config.json routines array: list each routine as id (schedule, on/off).
9. Read .claude-code-hermit/state/reflection-state.json for last_reflection timestamp and counters. Read .claude-code-hermit/state/alert-state.json for the most recent alert timestamp. Compute relative age ("2h ago", "45m ago").
10. Glob file counts: .claude-code-hermit/raw/** (excluding .archive/), .claude-code-hermit/compiled/**, .claude-code-hermit/raw/.archive/**.

Format the additional sections:

Proposals: N proposed, N accepted, N in_progress
Micro: N pending decision
Routines: morning-brief (08:00 daily), weekly-deps (09:00 mon, off)
Last: reflect 2h ago, heartbeat 45m ago
Reflect: N runs, N empty | judge: N accepted / N downgraded / N suppressed | output: N proposals, N micro | since YYYY-MM-DD
Knowledge: N raw, N compiled, N archived

• If a section has no data or the source file is missing, show — (not an error)
• Omit the Micro line entirely if there are zero pending micro-proposals
• Reflect line rules:
  • If counters is absent: Reflect: —
  • If total_runs is 0: Reflect: no runs yet (since YYYY-MM-DD)
  • Otherwise: full format as above; omit the | output: ... clause if both proposals_created and micro_proposals_queued are 0
• Total output (session + infrastructure) should stay under 12 lines