Dispatch Watcher Gemini

Dispatch Watcher Skill — Gemini CLI Adapter

This is the Gemini CLI-specific adapter for the shared agent-presence protocol. All shared logic (pre-flight, claim, lease, roster, takeover, presence event format) lives in .agent/skills/agent-presence/SKILL.md and scripts/agent-presence-helpers.sh — this file only contains Gemini-specific glue.

When this Gemini CLI session starts (or when explicitly triggered by @gemini #<num> in <DISPATCH_CHANNEL>), follow the workflow below.

Contract

• Input: environment variables AGENT_ID (starts with gemini-), GH_TOKEN_GEMINI_BOT, SLACK_BOT_TOKEN, ROSTER_MESSAGE_TS, ROSTER_CHANNEL_ID, REPO.
• Output: if a @gemini dispatch is pending, claim and execute it; otherwise return idle. start is emitted only AFTER pre-flight passes. end is emitted on session close via trap, with reason in EXIT_REASON.
• Slack Connectivity: Gemini CLI uses curl-based helpers in scripts/agent-presence-helpers.sh (which use SLACK_BOT_TOKEN) for both reading and writing to Slack. It does NOT use MCP tools.

Preconditions

1. Workspace name check (optional but recommended): When running multiple Gemini sessions in parallel, the seq for AGENT_ID is conventionally derived from a worktree directory named *-Gemini-<N>. Skip this check if running a single-session setup.

   # Portable extraction (no GNU grep -P / -oP): sed prints nothing on no match,
   # so WS_SEQ is genuinely empty (the `-z` test below works).
   WS_SEQ=$(basename "$PWD" | sed -n 's/.*Gemini-\([0-9][0-9]*\).*/\1/p')
   if [ -z "$WS_SEQ" ]; then
     # Single-session setup — fall back to a default seq
     WS_SEQ=1
   fi
   export AGENT_ID="gemini-${HOSTNAME:-host}-wsl-${WS_SEQ}"
   

2. Dependency check: Ensure shared protocol files are present before proceeding:

   if [ ! -f .agent/skills/agent-presence/SKILL.md ] || [ ! -f scripts/agent-presence-helpers.sh ]; then
     echo "Missing shared agent-presence dependencies (.agent/skills/agent-presence/SKILL.md or scripts/agent-presence-helpers.sh)" >&2
     exit 1
   fi
   

3. Tokens and repo: GH_TOKEN_GEMINI_BOT, SLACK_BOT_TOKEN, and REPO must be set. GH_TOKEN_BOT is resolved from GH_TOKEN_GEMINI_BOT:

   export GH_TOKEN_BOT="$GH_TOKEN_GEMINI_BOT"
   export REPO="<OWNER>/<REPO>"
   source scripts/agent-presence-helpers.sh
   

Workflow

0. Wire shutdown wrapper

Use the trap-based wrapper to ensure end event emission:

EXIT_REASON="boot-incomplete"

emit_presence_end_once() {
  local reason="${EXIT_REASON}"
  trap - EXIT INT TERM HUP
  emit_presence end "${reason}" || true
}

trap 'emit_presence_end_once' EXIT
trap 'EXIT_REASON="interrupted (${EXIT_REASON})"; emit_presence_end_once; exit 130' INT
trap 'EXIT_REASON="terminated (${EXIT_REASON})"; emit_presence_end_once; exit 143' TERM
trap 'EXIT_REASON="hangup (${EXIT_REASON})"; emit_presence_end_once; exit 129' HUP

1. Pre-flight check

Run pre_flight_check. If it fails, set EXIT_REASON="declined (pre-flight conflict)" and exit 0.

2. Emit start event

emit_presence start
EXIT_REASON="post-start"

3. Read dispatch channel for @gemini mentions

scripts/agent-presence-helpers.sh has no public helper for reading arbitrary channel history (it only exposes read_pinned_roster, which is hard-wired to the single pinned roster message). To scan the dispatch channel, call Slack's conversations.history directly with curl:

curl -sS "https://slack.com/api/conversations.history?channel=${DISPATCH_CHANNEL_ID}&limit=50" \
  -H "Authorization: Bearer ${SLACK_BOT_TOKEN}" \
  | jq -r '.messages[] | "\(.ts)\t\(.text)"'

Match each message text against: (^|[[:space:]])@gemini\b\s*#?([0-9]+)\b

Filter out messages older than 1h (compare each .ts epoch to now - 3600).

4. Claim and execute

For a matched <issue_num>, call claim_issue "$issue_num" quick.

If successful, follow the implementation workflow:

1. git checkout -b feature/issue-<num>-<slug>
2. Implement and verify per the project's Verification After Changes checklist
3. Open PR with Closes #<num>

5. Report completion

Emit the ready event, then record the worktree path on the roster. Note that emit_presence takes only <event> [<context>] (Slack messaging); the workspace_path lives on the roster, so set it via update_roster_chat_update:

emit_presence ready "PR #${PR_NUM} done"
update_roster_chat_update --event ready --workspace-path "$PWD"

Update EXIT_REASON="task-done (PR #${PR_NUM})".

Required env summary

export AGENT_ID="gemini-${HOSTNAME}-wsl-1"
export GH_TOKEN_GEMINI_BOT="ghp_..."
export SLACK_BOT_TOKEN="xoxb-..."
export REPO="<OWNER>/<REPO>"
export ROSTER_MESSAGE_TS="<ROSTER_MESSAGE_TS>"
export ROSTER_CHANNEL_ID="<DISPATCH_CHANNEL_ID>"
export GH_TOKEN_BOT="$GH_TOKEN_GEMINI_BOT"

source scripts/agent-presence-helpers.sh

Shared protocol references

• .agent/skills/agent-presence/SKILL.md — shared protocol contract
• scripts/agent-presence-helpers.sh — shared shell library
• .agent/skills/dispatch-watcher-claude/SKILL.md — Claude adapter
• .agent/skills/dispatch-watcher/SKILL.md — Codex adapter