Egregore Activity
SkillDev toolsShow Egregore team activity, pending handoffs, and open questions when the user invokes /activity or $activity, or asks what is happening in an Egregore checkout.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Egregore Activity skill
What this skill tells your AI
The instructions your AI receives, as published by egregore-labs/egregore in .codex/skills/activity/SKILL.md and read by ahel’s review.
Native Codex Egregore skill. Use this only inside an Egregore checkout.
Show the live team activity view without sending the user through Claude Code. The graph API is useful when reachable, but the filesystem and memory repo must remain enough to render a useful answer.
Flow
- Verify
bin/agent.shexists. If it does not, say this is not an Egregore checkout. - If the request is
activity done N,activity expire N, oractivity reopen N, fetchbin/activity-data.shto a temporary JSON file, mapNto the correspondinghandoffs_to_me[N-1].sessionId, and run:
bash bin/activity-action.sh <done|expire|reopen> "<session-id>"
Report the resulting topic and state. Never accept a raw session ID from the
visible command when a numbered activity item is available; the number keeps
the mutation tied to what the user just saw. In local mode, preserve the
script's explicit unavailable result.
3. Run bin/agent.sh sync.
4. Resolve the requested handle:
- If the user named a person, match it against
memory/people/*.md. - Otherwise use
.egregore-state.jsongithub_username, then git user fallback.
- Run
bash bin/activity-data.sh [handle]with network escalation and capture stdout as JSON. This command performs the graph API call internally; a normal sandboxed Codex shell can make a healthy graph look unreachable. - Classify the graph status:
node bin/codex-skill-render.mjs classify-graph --mode connected
If the network-enabled command still returns retry because graph_reason is
unreachable, render the filesystem fallback and mention that the graph service
was unreachable after a network-enabled attempt. If network escalation is denied
or unavailable, run the command without escalation, render the fallback, and
say Codex graph network access was not granted; do not claim the graph itself
was unreachable. Do not retry for missing_config, auth_error, or
server_error; render the clear reason instead.
- Render the card with:
node bin/codex-skill-render.mjs activity-card <json-file>
Paste the rendered card exactly in the visible response, preferably in a
text fenced block. Do not paraphrase the card into bullets unless the user
asks for a summary.
- Offer one compact focus prompt:
Handoffs- open pending handoffs.Questions- answer or inspect pending questions.Recent work- inspect sessions and wraps.Dashboard- switch to the personal dashboard.Done- stop.
Use structured Codex question tooling when it is available. Otherwise render a
numbered list plus Other: and wait for the user.
Rules
- Never show raw JSON.
- Do not narrate each command step unless something blocks the workflow; the final visible response should lead with the rendered card.
- Structured UX parity is required: preserve the rendered activity TUI card, no preamble, no prose-only replacement, and no raw collector output.
- Do not call
egregore-handoff. - Do not use Claude Code commands.
- Keep graph, pull request, and publish data best-effort; local memory remains the source of truth when connected services are down.
Signals
- GitHub stars
- 289
- Forks
- 21
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
activity- Source
- github.com/egregore-labs/egregore