routine-report

When to invoke this skill

Reads the run log of every routine in the folder and writes one weekly summary: what ran, what changed, what broke. Use when asked "what ran this week", "did my automations work", or as a routine of its own (Monday morning). Grit's skill; routines are made by /routine-setup.

The hard boundary

• From the logs only. A run happened if it has a row in its ## Run log (or the runtime's schedule history shows it). A scheduled run with no row is missed — no log row, never assumed fine.
• Quiet when nothing happened. No change and no failure → the report is written and no message is sent. A message that says "all good" every week trains the person to ignore the one that matters.
• Messages go to the person only. Through the channel they set up for localoy (or the runtime's own notification). A routine's "tell" field naming anyone else → the message to them is drafted in the report for the person's yes, never sent from here.
• Reports, does not repair. A broken routine is described with what the log shows; fixing it is a new run of /routine-setup (or /investigate for a site) after the person agrees.

What you read first

• Every PLAN-routine-*.md in the folder — its schedule, "changed means", "failed means", and ## Run log rows for the period.
• The runtime's schedule list and history, if it can be read — to catch routines with no plan, and runs with no log row.
• The last routine-report-*.md — so "still broken" can be told from "newly broken".

Period: the last 7 days unless the person names one.

Procedure

1. Load the inputs above.

Scratch for this run lives in .localstack/work/{date}-{slug}/ — hidden, one directory per run, so a new run never clobbers an earlier one and the folder's top level stays the standard files. Scratch is disposable; old run directories may be deleted freely.

2. Tally per routine: expected runs (from the schedule), logged runs, ok / changed / failed / missed counts, and the newest output file.

3. Pick out what matters: every changed row (what changed, quoted from the log), every failed or missed (what the log says went wrong; first seen date; still failing since the last report?).

4. Write routine-report-<yyyy-mm-dd>.md at the top of the working folder:

# Routines — week to YYYY-MM-DD

## Broke (N)
- <routine> — <failed/missed> <when> — <what the log says> — since <date>
## Changed (N)
- <routine> — <what changed> — <output file>
## Ran fine (N)
| Routine | Runs expected | Logged | ok |
## Not checked
- <routine with no log, schedule unreadable, …>

5. Message — only if Broke or Changed is non-empty. One short message to the person: the broken ones first, then the changes, each in one line, and the report's path. No message otherwise; say in the report "no message sent — nothing changed or failed".

Standard files. This folder is kept in files any agent already reads. Update them in place; never scatter output into new folders.

• AGENTS.md — create it if missing. localstack owns only the block between <!-- localstack:start --> and <!-- localstack:end -->; rewrite that block, never anything outside it. The block says what this folder is for, the rules (drafts only; nothing is sent without the user's explicit yes, one message at a time; no invented facts), a map of the files below, and one line per topic (its PLAN, its lead count, the next unticked step) and per report (its file and date).
• CHANGELOG.md — create it if missing (# Changelog). Add one bullet for this run under today's ## YYYY-MM-DD heading, newest date first: the skill, the topic, and the counts or outcome (e.g. - lead-search austin-dentists: 18 found, 3 skipped as already contacted).
• TODOS.md — create it if missing (# TODOs). Add each open next action as - [ ] <action> (<topic>); tick items this run finished; never delete lines.
• DESIGN.md — decisions meant to last (positioning, tone, channels to use or avoid). Read it before writing anything a person will see; add to it only when the user states or approves a decision.

For this step: the CHANGELOG line gives routines checked, ran / changed / broke counts, and whether a message was sent. One TODO per broken routine (- [ ] fix <routine>: <what broke> (routine-report)), not repeated if already open.

6. Report. The same three lines as the message, or "all routines ran, nothing changed".

Completion status. End the chat report with one of:

• DONE — completed, with the evidence named (files written, counts, URLs).
• DONE_WITH_CONCERNS — completed, and list each concern.
• BLOCKED — cannot proceed; say what blocked it and what was tried.
• NEEDS_CONTEXT — missing information; say exactly what is needed.

Quality bar

• Every "ran", "changed" and "broke" points at a log row.
• A missed run is never reported as fine.
• No message in a week where nothing changed or failed.