domains / agent-tooling / session-open / open

open

combined

The two copies are unrelated procedures that happen to share the name 'open'. The first is a session bootstrap for an 'aigent-OS' vault (preflight safety, heat index, session log, delegation, decision aging, runtime state). The second is a multi-mode research skill ('last30days') that routes to watchlist, briefing, history, or one-shot research modes. Neither is a reproduction of the other, and crowning one would hide a real alternative.

Repositories

Repository	Path
davepoon/buildwithclaude	`plugins/aigent-os/skills/open/SKILL.md`
LeoYeAI/openclaw-master-skills	`skills/last30days-skill/variants/open/SKILL.md`

Other copies

Role	Repository	Path
peer	davepoon/buildwithclaude	`plugins/aigent-os/skills/open/SKILL.md`
peer	LeoYeAI/openclaw-master-skills	`skills/last30days-skill/variants/open/SKILL.md`

Combined skill

# Session Open (Combined)

This skill covers two distinct 'open' workflows. Determine which applies based on the environment and user intent.

## Workflow A: aigent-OS Session Bootstrap

Use when `$AIGENT_ROOT` is set and the user is starting a normal working session in an aigent-OS environment.

1. **Preflight Safety Check**
   - Verify `$AIGENT_ROOT` is set and `vault/memory/` and `vault/daily/` exist.
   - If not, STOP and tell the operator to run `/setup`.
   - Run the read-only vault check script. If it prints `STOP`, halt.
2. **First-Run Detection**
   - Read `.aigent/state.json`. If `status` is not `ready`, run `/start`.
   - If JSON is missing but `.aigent/first-run-done` exists, create the JSON state.
3. **Unbanked-Session Recovery**
   - Compare `.aigent/last-close` with newest Session Captures in `vault/daily/`.
   - If newer captures exist, recover a summary into `vault/memory/SESSION_LOG.md` marked `(auto-recovered)`.
4. **Load Context**
   - Read `vault/memory/HEAT_INDEX.json` (if present) for prioritized reading.
   - Read latest daily note, session log, active priorities, delegation tracker.
   - Follow the graph into 3-5 relevant notes.
   - Check decision aging (30/60/90 days) and attention reconciliation.
   - Weekly: check skill gaps.
5. **Compute Runtime State**
   - Run `python3 "$AIGENT_ROOT/daemons/runtime/update-active-state.py"`.
   - Read `vault/memory/runtime/ACTIVE_STATE.json`.
6. **Output**
   - Greet the operator and surface only due decisions, material drift, stale items, dropped work, and open threads.

## Workflow B: last30days Research Skill

Use when the user invokes `last30` or `last30days` with a research, watchlist, briefing, or history request.

1. **Find Skill Root**
   - Locate `scripts/last30days.py` in standard skill directories.
2. **Load Context**
   - Read `variants/open/context.md` for user preferences.
3. **Route Command**
   - Parse the first argument:
     - `watch` → Read `references/watchlist.md`
     - `briefing` → Read `references/briefing.md`
     - `history` → Read `references/history.md`
     - *(other)* → Read `references/research.md`
4. **Execute Mode**
   - Follow the instructions in the corresponding reference file exactly.
   - Use SQLite database at `~/.local/share/last30days/research.db`.
   - Check API keys in `~/.config/last30days/.env` or environment variables.