Prj Report Daily

prj-report-daily

Overview

This skill is the daily reporter for the PR Jangler module. One invocation aggregates the persisted queue at {project-root}/_bmad-output/pr-workflow/state.json, every per-PR cache under {project-root}/_bmad-output/pr-workflow/prs/, and the last 24 hours of run-log lines; renders a multipart email (HTML with inline CSS plus a plain-text fallback); resolves SMTP credentials from 1Password via op read; sends through prj_smtp_host on prj_smtp_port with STARTTLS; archives the rendered body to {project-root}/_bmad-output/pr-workflow/reports/{YYYY-MM-DD}.md; and updates state.last_report_sent on success.

Act as a calm, no-bury operations summarizer. PleaseAdvise rises first. "Needs your eyes" is capped at 5 with an overflow pointer to the archive. If nothing happened, send the brief "all quiet" liveness email rather than skip; the user must hear from the bot every day.

Conventions

• Bare paths (e.g. scripts/run.py) resolve from the skill root.
• {project-root}/... resolves from the project working directory.
• Configuration variables live under [modules.prj] in {project-root}/_bmad/config.toml.
• All state I/O goes through prj-orchestrator's state_io module (imported via sys.path); this skill never touches state.json directly.

On Activation

Load configuration from {project-root}/_bmad/config.toml (the [modules.prj] block). Required keys:

• prj_repo (used in the subject + header for context)
• prj_email_to, prj_email_from
• prj_smtp_host, prj_smtp_port
• prj_smtp_creds_ref (1Password reference, resolved via op read for username and password fields)

If any required key is missing, exit with status misconfigured, log it, and do not touch SMTP. The runlog entry names the missing key.

Execute the report:

python3 scripts/run.py

That single script handles the full pipeline: aggregate, render, resolve creds, send, archive, persist last_report_sent, append run-log. Flags:

• --dry-run render and archive locally but skip SMTP and do not update last_report_sent
• --verbose emit progress diagnostics to stderr
• --project-root <path> override autodetect
• --now <iso> override "now" for deterministic testing/replay

See python3 scripts/run.py --help for full detail.

Architecture

Six Python modules, each with a single concern. The LLM activation context exists so the orchestrator's Skill-tool dispatch resolves cleanly when the daily-report cadence fires.

State and Files Touched

This skill writes to:

• {project-root}/_bmad-output/pr-workflow/state.json (only last_report_sent and report_send_failures)
• {project-root}/_bmad-output/pr-workflow/reports/{YYYY-MM-DD}.md (frozen archive of the rendered report)
• {project-root}/_bmad-output/pr-workflow/logs/{YYYY-MM-DD}.jsonl (append-only run-log)

This skill reads from:

• {project-root}/_bmad/config.toml
• {project-root}/_bmad-output/pr-workflow/state.json
• {project-root}/_bmad-output/pr-workflow/prs/{n}/ (review.md, decisions.log, verification.md, fix-plan.md, adversarial.md, implementation.md, meta.json)
• {project-root}/_bmad-output/pr-workflow/logs/*.jsonl (last 24 hours)

Aggregation Rules

Codified in scripts/aggregate.py; summarized here for activation review.

• needs-attention: PRs in PleaseAdvise plus PRs in FixImpl (an automated fix-PR was opened and is awaiting human eyes). Sorted by priority then by oldest phase_entered_at.
• in-progress: any PR in a non-terminal, non-PleaseAdvise phase (Discovered, Triaged, OverlapCheck, ReviewPending, Reviewed, CommentsTriage, ClaimVerify, FixPlan, AdversarialCheck).
• resolved-today: PRs whose phase transitioned to a terminal phase (ReadyToMerge, Blocked, Rejected, Archived) within the last 24 hours, sourced from last_action_at.
• no-change: PRs whose last_action_at is older than 24 hours and which are not already classified above. Count only (no per-PR detail in the email).
• "Needs your eyes" shortlist: top 5 across needs-attention. Overflow rendered as "+ N more, see archive."
• All-quiet: no needs-attention, no resolved-today, no in-progress activity in the last 24 hours (zero run-log entries with action=dispatch for per-PR skills). When all-quiet, the subject becomes [PR Jangler] {date}: all quiet and the body is a one-paragraph liveness confirmation.

Rendering Rules

• Inline CSS only. No <link>, no <script>, no external assets.
• HTML body is built by string.Template.substitute against assets/email-template.html. No Jinja2.
• Plain-text fallback is built by the same renderer and attached to a MIMEMultipart('alternative') with the HTML.
• After substitution, the renderer asserts no $ placeholder remains; an exception raises rather than send a broken email.
• Subject template (default): [PR Jangler] {date}: {n} need attention, {m} in progress. All-quiet override: [PR Jangler] {date}: all quiet.

Sending Rules

• Credentials: op read "{prj_smtp_creds_ref}/username" and op read "{prj_smtp_creds_ref}/password". Failures raise CredsError with a clear message; the runlog records status: "creds-error" and the script exits 4.
• Transport: smtplib.SMTP(host, port), starttls(), login(user, pass), sendmail(from, [to], message.as_string()), quit().
• On SMTP failure: append a runlog entry with status: "smtp-error" and increment state["report_send_failures"] (initialized to 0 if absent). last_report_sent is NOT updated. The script exits 5. The next heartbeat will retry. After 3 consecutive failures, the script also injects a synthetic system entry into state["prs"]["0"] (the digits-key sentinel; pr_number: 0) at phase PleaseAdvise, surfacing the failure to the user via the next successful report.

Persistence Rules

• Archive is written before SMTP send so a failed send still leaves a forensic copy.
• state.last_report_sent is updated ONLY after smtplib returns without raising. On retry the next heartbeat re-renders against the latest state (so the report content always reflects current reality, not a stale snapshot).
• report_send_failures counter resets to 0 on successful send and on the synthetic __system__ entry being created.

Non-Negotiables

• Never skip the daily email. If nothing happened, send the all-quiet liveness email.
• No buried critical items. PleaseAdvise rises to the top of the shortlist; the shortlist sits above all groupings in the body.
• Inline CSS only. No external assets in the email body.
• Atomic state writes via state_io.save_state. Direct writes to state.json are forbidden.

Verification

Smoke-test against a configured project (does not require SMTP; uses --dry-run):

python3 scripts/run.py --dry-run --verbose

Expected: exit 0, runlog entry with status: "dry-run", archive file present at reports/{YYYY-MM-DD}.md, state.last_report_sent UNCHANGED.

Run the unit tests (no SMTP, no op calls; all subprocess invocations mocked):

python3 -m unittest discover scripts/tests