open — Claude Code skill: what it does, how to install
This Claude Code skill loads workspace context and surfaces items that need attention at the start of a session. Running /open checks priorities and recovers unfinished work.
MIT ★ 26 updated 2026-11-16
npx skills add wrg32786/aigent-os --skill open -a claude-codeWhat open does
The skill starts a work session by reading notes, the session log, priorities, and runtime state, then reports only items that need attention. It checks aging decisions and drift between recent work and planned focus. Once a week, it scans for skill gaps and runs /skill-hunt on the oldest open one.
Before the main routine, it reads .aigent/state.json. If the status is not ready, it runs /start instead. It compares .aigent/last-close with recent Session Captures and writes recovered entries to vault/memory/SESSION_LOG.md. The skill is deprecated: /resume replaces it and runs automatically after a context clear.
The scripts require Node.js and Python 3, with AIGENT_ROOT pointing to the AIGENT root. The skill runs daemons/nightly-watchdog.mjs and daemons/runtime/update-active-state.py, and reads data under vault/ and .aigent/.
What it helps with
- recover session notes left unbanked
- find the next action in the session log
- surface aging decisions and blocked work
- compare recent activity with planned priorities
Example prompts for the agent
- Open a work session and show what needs attention.
- Run
/openand recover any unfinished session work. - Start the session by checking priorities, blockers, and open decisions.
How to install open
1Skills CLI (Vercel)
npx skills add wrg32786/aigent-os --skill open -a claude-code2Manually: put the files in place
~/.claude/skills/open/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/open/
1Skills CLI (Vercel)
npx skills add wrg32786/aigent-os --skill open -a codex2Manually: put the files in place
.agents/skills/open/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add wrg32786/aigent-os --skill open -a cursor2Manually: put the files in place
.cursor/skills/open/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install wrg32786/aigent-os skills/open2Manually: put the files in place
.github/skills/open/SKILL.md1Skills CLI (Vercel)
npx skills add wrg32786/aigent-os --skill open -a gemini-cli2Platform CLI
gemini skills install https://github.com/wrg32786/aigent-os --path skills/openManually: put the files in place
.devin/skills/open/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add wrg32786/aigent-os --skill open -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about open
How do I install open in Claude Code?
Run npx skills add wrg32786/aigent-os --skill open -a claude-code. The skill is installed at ~/.claude/skills/open/SKILL.md.
What does open do?
It loads workspace context and surfaces unfinished work, blockers, aging decisions, and priority drift. The skill is deprecated and superseded by /resume.
What does open need to run?
It needs Node.js, Python 3, and AIGENT_ROOT set to the AIGENT root. It reads files under vault/ and .aigent/.
Source text
SKILL.md · MIT
--- name: open description: "Boot a session, load vault context, and surface what matters. Use when opening a session, saying good morning, or starting work." trigger: /open status: DEPRECATED by the two-verb lifecycle — see skills/resume/SKILL.md. /resume absorbs /open; the post-clear boot runs it automatically via daemons/resume-verb.mjs. Kept for manual invocation during the transition; not deleted. --- # Session Open > **Deprecated.** This skill is superseded by [[skills/resume/SKILL.md|/resume]] (docs/two-verb-lifecycle.md). New installs should reach for `/resume`; this file stays for anyone still invoking `/open` by habit during the transition. Run this at the start of every normal working session. ## Nightly alert surface Before normal orientation, run the local watchdog with the same configured time zone and cutoff used by the pass: ```text node daemons/nightly-watchdog.mjs --root <aigent-root> ``` Surface every active named alert from `vault/memory/runtime/NIGHTLY_ALERTS.jsonl`. The append-only ledger and stderr are the default delivery channels; no external coordination service is required. If the ledger is unreadable, report `ALERT_LEDGER_INVALID` and continue the resume path. An alert-read failure must never suppress session orientation or make the hook indistinguishable from "not installed." ## First-run detection Before the normal protocol, read `.aigent/state.json`. - If `status` is not `ready`, run `/start` instead of the normal open protocol. - If the JSON state is missing but `.aigent/first-run-done` exists, treat the installation as ready and create the JSON state as a compatibility migration. - Do not inspect natural-language placeholders in `system/00_identity.md`. Product state belongs in machine-readable state, not in whether a sentence happens to survive an edit. ## Protocol **Unbanked-session recovery:** Compare `.aigent/last-close`, if present, with the newest Session Captures in `vault/daily/`. If newer captures exist, recover a short summary into `vault/memory/SESSION_LOG.md`, mark it `(auto-recovered)`, and continue. State the recovery in one neutral line. 1. **Load the heat index** from `vault/memory/HEAT_INDEX.json` when present. Use `hot_top_20` as the prioritized reading list. If missing, use steps 2 through 6 normally. 2. **Read the latest daily note** in `vault/daily/`. 3. **Read the session log** at `vault/memory/SESSION_LOG.md` and extract the newest next action and open threads. 4. **Read active priorities** at `vault/memory/ACTIVE_PRIORITIES.md`. 5. **Follow the graph** from the latest daily note and open threads into the three to five most relevant project, concept, person, or agent notes. 6. **Check delegation** at `vault/memory/DELEGATION_TRACKER.md` for blocked, stale, or review-ready work. 7. **Decision aging**: compare `vault/memory/DECISION_LOG.md` and `vault/memory/DECISION_OUTCOMES.md`. Surface at most three unresolved decisions around 30, 60, or 90 days old. 8. **Attention reconciliation**: compare the last seven days in `vault/daily/` with intended focus in `vault/memory/ACTIVE_PRIORITIES.md`. Surface material drift only. 9. **Skill gap scan**: once per week, inspect `memory/SKILL_GAPS.md`. If an open gap is older than seven days, run `/skill-hunt` on the oldest one and update the ledger if resolved. 10. **Compute runtime state**: ```bash python3 "$AIGENT_ROOT/daemons/runtime/update-active-state.py" ``` 11. **Read runtime state** at `vault/memory/runtime/ACTIVE_STATE.json`. Use `next_valid_action`, `blocked_items`, active reflexes, and mode to orient the session. 12. **Reconcile weekly**: on Monday or after seven days without a reconciliation, run `/reconcile` and surface contradictions only. ## Output Give a brief greeting, then surface only: - Due decision-aging prompts. - Material attention drift. - Stale or blocked items. - Work dropped from the previous next action. - Open threads requiring attention now. If nothing needs flagging, greet the operator and let them drive. Do not dump a full priority report merely because the files exist.
© wrg32786 · MIT · wrg32786/aigent-os