Dailybot Agent Skill

The Dailybot Agent Skill is a single cross-agent skill pack that installs into every major AI coding CLI with one command. Once installed, your agent has the same verbs a human teammate has — report progress, complete check-ins, send an email, kudos a colleague, post on Slack or Teams, submit a form, or announce its own health. Every sub-skill maps 1:1 to a public Dailybot API endpoint. This page is the technical reference; for the plain-English pitch, read /skills/agent-skill.

Install

Prerequisites: dailybot-cli >= 1.10.0 as a baseline —>= 1.17.2 is required for the authoring surface (check-ins/forms create + config, report channels). Theagent-skill router installs the CLI on first use fromhttps://cli.dailybot.com/install.sh, verified against a SHA-256 checksum. The skill pack itself is pinned atDailybotHQ/agent-skill >= 1.8.3.

Recommended — any AI CLI

Install the CLI baseline, then add the skill to your project:

curl -fsSL https://cli.dailybot.com/install.sh | bash
npx skills add DailybotHQ/agent-skill

Claude Code · Cursor · OpenAI Codex CLI · Gemini CLI

npm-based agents read .agents/skills/ directly. One install is enough:

npx skills add DailybotHQ/agent-skill

GitHub Copilot · Cline · Windsurf

Any harness that reads .agents/skills/ works. Clone the skill into place:

git clone https://github.com/DailybotHQ/agent-skill \
  .agents/skills/dailybot-agent

Manual

git clone https://github.com/DailybotHQ/agent-skill
cd agent-skill && ./setup.sh

Filesystem layout

Once installed, the skill lives inside your project's.agents/skills/dailybot-agent/. Per-repo policy and identity live under .dailybot/. Every file is plain text — no binaries, no bundled Node modules. Installing is a copy operation; uninstalling is rm -rf .agents/skills/dailybot-agent/.

<repo>/
├── .agents/
│   ├── skills/
│   │   └── dailybot-agent/
│   │       ├── SKILL.md          # top-level router
│   │       ├── report/SKILL.md
│   │       ├── checkin/SKILL.md
│   │       ├── chat/SKILL.md
│   │       ├── kudos/SKILL.md
│   │       ├── forms/SKILL.md
│   │       ├── teams/SKILL.md
│   │       ├── health/SKILL.md
│   │       ├── email/SKILL.md
│   │       ├── conversation/SKILL.md
│   │       ├── ask/SKILL.md
│   │       ├── messages/SKILL.md
│   │       ├── workflow/SKILL.md
│   │       └── channels/SKILL.md
│   └── settings.json             # optional harness hooks (Claude Code)
├── .cursor → .agents             # symlink (Cursor reads hooks.json here)
├── .claude → .agents             # symlink for legacy Claude Code paths
└── .dailybot/
    ├── profile.json              # per-repo identity + policy (commit)
    └── disabled                  # opt-out flag (commit or keep local)

Authentication

The skill is a thin wrapper over the dailybot CLI. Auth is inherited from whichever credential the CLI resolves first — both credentials have identical scopes (parity is a written guarantee, no scoped-down surprises in CI).

Precedence: DAILYBOT_API_KEY wins when both are set. Verify the resolved credential with:

dailybot status --auth

Full details: Authentication.

Sub-skill reference

The pack exposes thirteen sub-skills, one per public Dailybot verb. Each is a thin wrapper over the corresponding dailybot CLI command — so anything you can do with the skill you can also do with the CLI directly.

report — progress updates

Wraps dailybot agent update. Sends a standup-style update as the agent, crediting the running human as co-author whenever a session token is present. Rules: 1–3 sentences, WHAT + WHY, English only — never file paths, commit hashes, or branch names.

dailybot agent update "Shipped the pricing refresh — Advanced is now the recommended tier." \
  --name "Cursor agent" \
  --metadata '{"model":"composer-2.5"}'

checkin — standups + authoring

dailybot checkin list          # pending today
dailybot checkin show <uuid>   # config + questions
dailybot checkin complete <uuid>
dailybot checkin history <uuid>

Authoring commands (CLI >= 1.17.2): dailybot checkin create,dailybot checkin edit, dailybot checkin reset.

chat — cross-platform bot messages

Same command surface for Slack, Microsoft Teams, Discord, and Google Chat — the connected platform is resolved server-side, so the skill never has to know which chat vendor your team uses.

dailybot chat send --to @user "Text goes here"
dailybot chat send --channel "#team" --text "Text goes here"
dailybot chat update <message_uuid> --text "Edited text"

kudos — recognition

dailybot kudos give --to @teammate --message "Good review on #199"
dailybot kudos give --team Engineering --message "Shipped the release form v9"

forms — response lifecycle + authoring

dailybot form list                                # forms visible to you
dailybot form get <uuid> --json                   # full payload (questions + states)
dailybot form submit <uuid> --content '{...}'     # create a response
dailybot form transition <uuid> <rid> qa          # advance workflow state
dailybot form response get <uuid> <rid> --json    # read a response

Authoring commands (CLI >= 1.17.2): dailybot form create,dailybot form update, dailybot form delete. See the Forms API reference for the underlying HTTP surface.

teams — resolve names to UUIDs

dailybot team list
dailybot team get Engineering       # case-insensitive
dailybot team get <uuid>

Other sub-skills call teams internally when they need to translate a human-readable team name — you rarely invoke it directly.

health — heartbeat + inbox

A successful health check delivers any pending messages back to the agent, so long-running agents poll for team instructions with the same call that keeps them visible in the standup.

dailybot agent health --ok --message "All good"
dailybot agent health --fail --message "DB unreachable"
dailybot agent health --status --name "Claude Code"

email — send an email

Consent-gated: the recipient must be pre-confirmed in the agent profile. Bounces come back as agent messages so the skill can react.

dailybot agent email send \
  --to [email protected] \
  --subject "Weekly summary" \
  --body "..."

conversation — group DMs

Opens or resumes a group DM thread on the connected chat platform. If the same participant set already has an open thread, the CLI reuses it instead of creating a duplicate.

dailybot conversation open -u @jane @bob -m "Deploy report attached"
dailybot conversation open -u @jane              # reuses existing group if same people

ask — AI assistant

One-shot or interactive chat with the Dailybot AI. Supports structured JSON output for programmatic consumption, or a full-screen TUI when invoked without arguments.

dailybot ask "Summarize this week's standup responses"
dailybot ask --json                               # structured output
dailybot ask                                      # interactive TUI mode

messages — agent inbox

Retrieves pending messages addressed to the agent by team members. Useful for long-running agents that need to poll for human instructions between health checks.

dailybot agent message list --name "Claude Code" --pending

workflow — inspect workflows

Read-only listing and inspection of configured Dailybot workflows. Agents use this to understand what automations are already running before proposing changes.

dailybot workflow list
dailybot workflow get <uuid> --json

channels — report channels

Lists the report channels (Slack, Teams, Discord, Google Chat) visible to the authenticated user. Agents call this to discover where to deliver reports and messages.

dailybot channels list --json

Repo-level .dailybot/profile.json

Committed identity + policy for the repo. No secrets — credentials always live in DAILYBOT_API_KEY or the local session token.

{
  "agent_name": "cursor-agent",
  "default_metadata": {
    "model": "composer-2.5"
  },
  "report": {
    "min_interval_minutes": 20,
    "nudge": {
      "unreported_commits_threshold": 3,
      "sustained_work_minutes": 45
    }
  }
}

• agent_name — signs every report so a mixed-agent repo (Cursor + Claude Code) is distinguishable in the standup feed.
• default_metadata — merged into every dailybot agent update payload.
• report.min_interval_minutes — skill-side rate-limit for reports. The server also enforces a 30-minute floor.
• report.nudge.* — thresholds for the local hook reminders (unreported commits, sustained work).

Silencing per repo

Drop an empty file to disable every skill hook + telemetry for one repo. Commit it if you want the silence to travel with the repo; keep it local if it's just for you.

touch .dailybot/disabled

Session hooks

The skill ships deterministic harness hooks in two flavors — both are best-effort, local-only, and never call the network. The full lifecycle is: session-start → activity → post-commit → stop.

On session start the hook injects any login nudge + a summary of unreported work from prior sessions. On file edits it records signals into a local per-repo ledger. On post-commit it fires a stronger commit-level signal for the report nudge heuristic. On stop it decides — from the ledger — whether to inject a "consider sending a Dailybot report" reminder into the agent's context. If the reminder fires and the agent has nothing to say,dailybot hook dismiss snoozes it for an hour.

Full spec: hooks.md in the skill repo.

Troubleshooting

• dailybot: command not found — the router fell back to a manual install. Run curl -fsSL https://cli.dailybot.com/install.sh | bash and re-run the sub-skill.
• 401 Unauthorized from any sub-skill — no auth was resolved. Run dailybot login, or export DAILYBOT_API_KEY before invoking the agent.
• Sub-skill doesn't appear in the harness — .agents/skills/dailybot-agent/ isn't on disk. Re-run the install command.
• Rate limit — you exceeded the per-agent report interval. Lower the frequency, or bump report.min_interval_minutes in .dailybot/profile.json.

Versioning

The agent-skill router publishes a manifest for each release; the CLI bootstrap pins itself against that manifest so a stale skill pack cannot pull in an unexpected CLI.

Go deeper

• Marketing overview at /skills/agent-skill — sub-skill catalog, features, FAQ
• Skill repo on GitHub
• Dailybot CLI reference
• Authentication
• API reference