Dailybot Report

Dailybot Progress Report

You help developers keep their team informed. After completing tasks or making broad edits, you send a short progress update to Dailybot. The update reads like a human giving their daily standup — no one should be able to tell it came from an agent.

Agent Compatibility

This skill works across all major AI coding agents. Execution paths differ slightly but the logic is identical.

When natural language triggers apply, the skill activates on phrases like: "report this to Dailybot", "send a Dailybot update", "let my team know what we built".

Step 0 — Optional Auto-Activation Setup (First Run Only)

Auto-activation makes Dailybot reporting fire automatically in future sessions without the developer invoking the skill each time. It is opt-in. Writing to a global agent config file (CLAUDE.md, AGENTS.md, etc.) affects every future session in every repository, including unrelated work — so the developer must explicitly approve it the first time.

There are two independent opt-ins, and you should consider both every first run:

• Step 0a — Prompt trigger (works on every agent): a standing instruction in the global config file.
• Step 0b — Hook enforcement (harnesses with a hook system): deterministic, harness-driven reminders.

Check each one separately. Having the trigger installed does NOT mean hooks are installed — a returning developer who accepted the trigger in an older version still needs the hook offer. Never skip Step 0b just because Step 0a is already done; only skip an opt-in whose own check shows it is present.

Step 0a — Check if the prompt trigger is already installed

Identify which agent you are and run the corresponding check:

If output says "installed" — the prompt trigger is set; skip the rest of Step 0a and go to Step 0b (do not skip to Step 1).

If missing — ask before creating

Read triggers.md for the exact file path and content for the current agent.

If DAILYBOT_AUTO_YES=1 is set in the environment, treat consent as already given — write the trigger block straight away and skip the prompt below. Otherwise, show the developer this prompt before writing anything:

"Dailybot reporting can run automatically after every task or broad edit, across all your future sessions. To set that up, I'd add the block below to <file path>:

<exact content from triggers.md, including the
dailybot-auto-activation markers>

This affects every repository you open with this agent. To uninstall later, delete the marked block — see the Uninstall section in the README.

Should I add it now? (yes / no / not now)"

• Yes — write the trigger block exactly as shown, including the dailybot-auto-activation opening and closing markers from triggers.md, then continue to Step 0b.
• No / not now — skip the trigger for this session and continue to Step 0b (hooks are a separate opt-in worth offering even if the trigger was declined). Do not re-ask about the trigger in the same session.

If the developer accepts, briefly confirm:

"Added. From your next session onward, Dailybot reporting fires automatically. Remove the marked block in <file path> to disable."

Then continue to Step 0b (do not skip to Step 1 — hooks are a separate opt-in).

Step 0b — Hook enforcement (recommended)

Trigger blocks are advisory — the model can forget them in long sessions. The CLI ships a dailybot hook command group that the harness itself invokes at session start, after file edits, and at the end of every turn, making the reminders deterministic. This is what makes reporting truly autonomous, so offer it whenever it is available — including to developers who already accepted the Step 0a trigger in a previous session.

Check whether hooks apply, and whether they are already installed

1. CLI version. Run dailybot --version. If it reports below 3.9.0 (or the CLI is absent), skip Step 0b silently and continue to Step 1 — the pack baseline is dailybot-cli >= 3.9.0, and the Step 0a trigger alone still works on older installs.

2. Harness support. If the current harness has no lifecycle-hook system (e.g. Cline today), skip Step 0b silently — Step 0a covers it.

3. Already installed? Check whether the hook config already references dailybot hook, and skip to Step 1 if so:

   Agent	Check
   Claude Code	grep -rq "dailybot hook" ~/.claude/settings.json .claude/settings.json 2>/dev/null && echo "installed"
   Cursor	grep -rq "dailybot hook" ~/.cursor/hooks.json .cursor/hooks.json 2>/dev/null && echo "installed"
   Codex	grep -rq "dailybot hook" ~/.codex/hooks.json .codex/hooks.json 2>/dev/null && echo "installed"
   Copilot	grep -rq "dailybot hook" .github/hooks/ 2>/dev/null && echo "installed"
   Gemini CLI	grep -rq "dailybot hook" ~/.gemini/settings.json .gemini/settings.json 2>/dev/null && echo "installed"

   If output says "installed" — hooks are already set; continue to Step 1.

If missing — ask before creating

Read hooks.md for the current harness's exact config file path, content, merge rules, and uninstall path. The same consent rules as Step 0a apply: show the file path and exact content before writing, merge (never overwrite) existing config, and DAILYBOT_AUTO_YES=1 counts as consent. A repo-level config (.claude/settings.json, .cursor/hooks.json, …) committed to the repository gives the whole team autonomous reporting on clone — offer that when the developer is in a shared project.

Once hooks are wired, consider the repo's report cadence. For research- or docs-heavy repos where much of the valuable work never lands as a commit, add "report": {"mode": "continuous"} to the committed .dailybot/profile.json so non-commit work (research, analysis, design docs, plans) is nudged sooner (a lower interval and turn threshold). See hooks.md § Per-repo controls.

If the developer declines, continue to Step 1 and do not re-ask in the same session. The Step 0a trigger still provides (probabilistic) coverage.

Responding to injected Dailybot reminders

When a hook injects a reminder ("commits have landed since the last report…" / "sustained work without a progress report…") into your context:

• Meaningful unit complete (including non-commit work — research, analysis, documents) → compose and send the report now (Steps 3–7).
• Nothing significant / still mid-stream → run dailybot hook dismiss to snooze for an hour. Either report or dismiss — never ignore the reminder silently.

Step 1 — Verify Setup

Read and follow the authentication steps in ../shared/auth.md. That file covers CLI installation, login, API key setup, and agent profile configuration.

If auth fails or the developer declines, skip reporting entirely and continue with your primary task.

Step 2 — Choose Execution Path

command -v dailybot

• CLI found → Step 4A
• CLI not found → Step 4B (see ../shared/http-fallback.md for the base curl patterns)

Both paths produce identical results. Prefer CLI — it handles auth and retries automatically. Fall back to HTTP in sandboxed environments, CI, or containers where the CLI cannot be installed.

Step 3 — Detect Context

3a. Run the bundled script

Find context.sh in the shared/ directory of the Dailybot skill pack and run it:

bash ~/.cursor/skills/dailybot/shared/context.sh

Adjust the base path to match your agent's skill directory (e.g., ~/.claude/skills/dailybot/, ~/.codex/skills/dailybot/).

Outputs JSON: {"repo":"...","branch":"...","agent_tool":"...","agent_name":"..."}. Capture it.

3b. Manual fallback (any agent)

See the header comments in ../shared/context.sh for the manual fallback commands:

# repo name
git remote get-url origin 2>/dev/null | sed 's|.*/||;s|\.git$||'

# branch
git branch --show-current 2>/dev/null

If git commands fail, use the current folder name for repo and "unknown" for branch.

Assemble metadata — include your model identifier:

{
  "repo": "<repo_name>",
  "branch": "<branch_name>",
  "agent_tool": "<your_tool>",
  "agent_name": "<your_name_or_profile_name>",
  "model": "<your_model>"
}

Model identifier examples: "claude-sonnet-4-6", "o3", "gemini-2.5-pro", "gpt-4o".

Step 4A — Send Report via CLI

Use this path when command -v dailybot succeeds.

Timeout: Allow at least 30 seconds for CLI commands to complete. Do not use a shorter timeout.

Pre-flight (mandatory) — respect the repo profile

Before constructing the command, do the repo-profile pre-flight from the router: ../SKILL.md § Mandatory pre-flight. Full procedure in ../shared/repo-profile.md. One-liner detection:

ROOT="$(git rev-parse --show-toplevel 2>/dev/null || pwd)"
[ -f "$ROOT/.dailybot/profile.json" ] && cat "$ROOT/.dailybot/profile.json"

The examples below show --name and --metadata for the case where the repo has no .dailybot/profile.json (no repo profile, no pinned identity). If a repo profile exists, omit the flags it already provides — see the per-key table in the router. The Plain report (with repo profile) block at the end shows what each example collapses to in that case.

Flag reference

Plain report — no repo profile

dailybot agent update "<message>" \
  --name "<agent_name>" \
  --metadata '<metadata_json>'

Rich report (multiple deliverables) — no repo profile

dailybot agent update "<message>" \
  --name "<agent_name>" \
  --json-data '<structured_json>' \
  --metadata '<metadata_json>'

Milestone report — no repo profile

dailybot agent update "<message>" \
  --name "<agent_name>" \
  --milestone \
  --json-data '<structured_json>' \
  --metadata '<metadata_json>'

With repo profile (the common case)

When .dailybot/profile.json sets name (and optionally default_metadata), the same three patterns collapse to:

# Plain
dailybot agent update "<message>" \
  --metadata '<metadata_json_minus_profile_keys>'

# Rich
dailybot agent update "<message>" \
  --json-data '<structured_json>' \
  --metadata '<metadata_json_minus_profile_keys>'

# Milestone
dailybot agent update "<message>" \
  --milestone \
  --json-data '<structured_json>' \
  --metadata '<metadata_json_minus_profile_keys>'

Where <metadata_json_minus_profile_keys> is your inline metadata with every key already set by default_metadata removed (the CLI shallow-merges them server-side; passing them inline is at best redundant, at worst overrides them silently).

Worked example — given {"name":"CLI","default_metadata":{"repo":"cli"}} at repo root, a plain report becomes:

dailybot agent update "Implemented X to fix Y." \
  --metadata '{"model":"claude-opus-4-7"}'

(no --name, no repo in --metadata — both come from the profile).

Using a named profile (no repo profile)

If a non-default global profile was configured during auth setup AND there is no repo profile pinning it:

dailybot agent --profile <profile_name> update "<message>" ...

If a repo profile sets profile, omit --profile from the command line.

Step 4B — Send Report via HTTP API

Use this path when the CLI is unavailable. Requires curl. See ../shared/http-fallback.md for the base patterns.

Base URL: https://api.dailybot.com Auth header: X-API-KEY: $DAILYBOT_API_KEY

Plain report

curl -s -X POST https://api.dailybot.com/v1/agent-reports/ \
  -H "X-API-KEY: $DAILYBOT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agent_name": "<agent_name>",
    "content": "<message>",
    "metadata": <metadata_json>
  }'

Rich report

curl -s -X POST https://api.dailybot.com/v1/agent-reports/ \
  -H "X-API-KEY: $DAILYBOT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agent_name": "<agent_name>",
    "content": "<message>",
    "structured": {
      "completed": ["Deliverable 1", "Deliverable 2"],
      "in_progress": ["Item still being worked on"],
      "blockers": []
    },
    "metadata": <metadata_json>
  }'

Milestone report

curl -s -X POST https://api.dailybot.com/v1/agent-reports/ \
  -H "X-API-KEY: $DAILYBOT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agent_name": "<agent_name>",
    "content": "<message>",
    "is_milestone": true,
    "structured": {
      "completed": ["..."],
      "in_progress": [],
      "blockers": []
    },
    "metadata": <metadata_json>
  }'

Response

A successful request returns 201 with the created report, including a url field that points to where the report landed in Dailybot:

{
  "id": "<uuid>",
  "url": "https://app.dailybot.com/agents/report/<uuid>",
  "...": "..."
}

Capture that url and surface it when you confirm (see Step 7). The CLI prints the same link automatically as a View: line; the HTTP 201 body above always carries the url too — read it from there.

Step 5 — Trigger Check

Send a progress report after each of the following:

1. Task completed — You finished a discrete task or subtask (e.g., a todo item, a user-requested change, a bug fix, a feature).
2. Broad edit — You modified 3 or more files in a single batch of edits.

Once auth is set up and (if applicable) the auto-activation trigger is in place, you do not need to re-confirm each individual report — the developer already opted in. Don't second-guess whether the work is "significant enough"; if either condition above is met, send the report.

If the developer explicitly asks you to report, always report regardless of these conditions.

What to include

Every report must answer two questions:

• What changed — concrete outcomes, not file paths
• Why — the purpose or motivation behind the change

Aggregate related work

If you completed multiple related changes, combine them into one report. Don't send 3 reports for parts of one feature — send one report covering the whole thing.

For detailed guidelines, see significance.md.

Step 6 — Compose the Report

The Human-First Principle

The developer directed the effort, made the decisions, and is accountable for the result. Your report reflects their accomplishment.

Golden rule: The reader should never be able to tell whether a human or an agent wrote the report.

• Never say "Agent completed...", "Claude did...", "I implemented...", "The AI built..."
• Focus on outcomes: what was accomplished and why it matters
• Always write in English, regardless of conversation language
• 1–3 sentences maximum, past tense

Report type

Plain report — single bug fix, small feature, one-off task. Message + metadata only.

Rich report — multi-deliverable feature, major refactor, complex task. Message + structured data + optional milestone flag.

Structured data format

{
  "completed": ["Deliverable 1", "Deliverable 2"],
  "in_progress": ["Item still being worked on"],
  "blockers": ["Issue preventing progress"]
}

Each item: concise, human-readable string. Empty arrays are fine.

Milestone flag

Mark a report as a milestone only when the top-level task is fully completed — all subtasks are done and the entire requested piece of work is wrapped up. Individual subtask completions are regular reports, not milestones. If the developer explicitly asks for a milestone, always honor it.

Co-authors

Do not add --co-authors by default — Dailybot automatically credits the authenticated developer. Only add if the developer explicitly asks to credit someone else. Never guess email addresses.

Forbidden in report messages

For writing templates by work type, see writing-guide.md. For side-by-side examples, see examples.md.

Step 7 — Confirm

After the command runs:

• Success — briefly confirm what was reported, and include the placement link the response returns (the url field, printed by the CLI as View:, and always present in the HTTP 201 body) so the developer can jump straight to where it landed. Example: "Reported to Dailybot: Built the notification preferences system with full test coverage — view it at https://app.dailybot.com/agents/report/." If no url is surfaced (an older backend), just confirm the report without a link.
• Failure — warn briefly. Do not retry in a loop. Suggest dailybot status --auth for auth issues, or dailybot logout + dailybot login if the session seems stale.
• Skipped — say nothing. Complete silence is the correct response.

Non-Blocking Rule

Reporting must never block your primary work. If the CLI is missing, auth fails, the network is down, or the command errors:

1. Warn the developer briefly
2. Continue with the primary task
3. Do not retry automatically
4. Do not enter a diagnostic loop

Additional Resources

• triggers.md — auto-activation trigger templates for each supported agent
• hooks.md — deterministic hook enforcement: per-harness configs, reminder handling, dismiss
• significance.md — when to report and when to stay silent, with edge cases
• writing-guide.md — writing templates by work type, action verbs, rate limiting
• examples.md — 15 side-by-side good vs bad comparisons
• ../shared/auth.md — authentication setup (CLI login, API key, agent register)
• ../shared/context.sh — automated context detection
• ../shared/http-fallback.md — HTTP API fallback patterns
• ../shared/dashboard-urls.md — dashboard URL catalog (report detail pages, agent analytics, etc.)
• Live API spec: https://api.dailybot.com/api/swagger/
• Full agent API skill: https://www.dailybot.com/skill.md