agentleFS
ilya-kozyrevilya-kozyrev/delamain/skills/progress-report/SKILL.md

progress-report

Skill5 starsChanged 36 days agoMITView on GitHub

Build a one-page infographic status of a stage, phase or project for its owner, team or client — how much is done by the scope's own definition of done, what changed since the last report, per-track items, figures with their sources, what is left and who it waits on — as an HTML page from the bundled template. Use for “status with an infographic”, “progress report”, “how much is left”, “статус с инфографикой”, “сколько осталось по фазе”, “отчёт о прогрессе для команды”. Not for live agent status (skill `status`) or a shift handoff (skill `handoff`).

10 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Files
SKILL.md

Progress report

The page answers, for a reader who was not in the work: how much is done, what changed, what is left, what needs me. Counts and the bar are computed from a list of items, so they cannot disagree; anything typed by hand carries a source. Development methods differ — many MRs or one per phase, a tracker or a checklist, production or a release tag or no code at all — so the scope's own definition of done drives every step, not a default.

  1. Pick the shape before collecting anything. Answer in a few lines of text now, and offer the page afterwards, when the reader needs it within ~30 minutes, the reader is another hub or agent, or the stage is small, idle or closed and its whole status fits in two lines. The page pays off for work with several items, a reader outside the work, or a recurring request (the next report reuses this one).
  2. Fix the scope, the item, and what done means. The item is whatever the work is tracked by: tracker issues, a task list in an issue, the checklist of a spec or of a single phase MR, plan steps, register entries (P-, Q-), acceptance criteria, measurements. Write done_means in the reader's words: “on production”, “merged and tagged”, “ticked in the phase checklist, deploys at the end of the phase”, “accepted by the owner”, “it runs locally”. The total is the written scope. With no written list, build it from the plan, registers and journal, say so in a note with the date, and do not pad it; if the plan ends before the stage's goal, add the goal as an item.
  3. Establish each item's status from the evidence its done_means points at: deploy or release tags, merged changes, MR/PR state and CI, ticked boxes in the spec or issue, the tracker, a run or test you can repeat, data you can re-read, the owner's recorded acceptance, the hub journal, handoffs, ask summary, live agents (<plugin-root>/bin/agent-top --json, skill status). Resolve the plugin root from PLUGIN_ROOT, CLAUDE_PLUGIN_ROOT or this skill's location; delegate wide reading by the session's rules. Record the sources.
  4. Map to the statuses. done meets done_means; review is finished and awaiting a check (review, CI, test window, deploy) — when one MR carries the whole phase, say in the label or a note that it is checked as one; active is being done now; planned is not started; blocked waits on someone outside the reader's control (another team or hub, a vendor, a delivery); decision waits on the reader or owner; unknown could not be verified — never done; dropped left the scope (label it, e.g. “→ next phase”; it is excluded from the total). An item's label refines its pill without changing its status.
  5. Fill the template. Copy <plugin-root>/skills/progress-report/resources/template.html, set <title> (two to four words) and replace only the JSON in <script id="report">:
    • lang, eyebrow (scope · date and time with zone), headline with {count}, count_format (any status key, total and dropped are available, also in lede and track_count_format), done_means, lede;
    • show_bar: false when progress is not a share of items — a measurement, an incident window, a single deliverable; the headline then states the result in words and metrics carry the figures;
    • metrics[]: label, value, source (where it was read, and when) — money, scores, times, sizes; a figure without a source is flagged on the page;
    • statuses: pill and legend labels in the reader's language;
    • tracks[]: title, summary, items[] of optional code, title, status, optional label; optional wide;
    • previous: date and items copied from the previous report's saved JSON (each with code or title and status), with changes_format ({date}, {done_delta}, {added}, {removed}): the page computes what moved and how the scope changed, so a grown total does not read as regression;
    • road[]: steps with kind now / soon / next / blocked / decision, when, title, items[];
    • notes[]: gaps the reader must know (no exit criterion, a reconstructed list, stale roadmap); foot: sources. Text is plain; backticks render as code. Write for the reader: outcomes, not process; leave out what the reader must not see (another person's pending decisions, internal topology) rather than relying on the page being private.
  6. Check before showing. Render the page (built-in browser, or at least python3 -m json.tool on the JSON) and confirm: no red problem banner; the total is the scope from step 2; every typed number in the lede, labels and metrics was re-read from its source in this run; it reads at phone width.
  7. Deliver where the reader is. Save the HTML and its JSON side by side (the JSON feeds the next previous): <stage>/coordinator/progress-<YYYY-MM-DD>.html in a hub stage, otherwise next to the work. Then: for the owner in Claude Code, a private page via the Artifact tool; for a reader in a tracker issue or chat, a phone-width screenshot or a short markdown summary with the file; for anyone else, the self-contained HTML file (it renders offline with system fonts). Money, personal data, production topology or a stage rule against publishing — ask the owner before any hosted copy. Sending anything outside the company is the owner's call.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.