agentleFS
rajasdevelrajasdevel/outcomebound/.agents/skills/sync-tracker/SKILL.md

sync-tracker

Skill0 starsChanged 15 days agoApache-2.0View on GitHub

Use when reading or writing a card in a tracker outside the project's ticket store. A write needs a grant naming its kind.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read through the harness's connector, a command-line tool the person has signed in to, or an export the person gives you.

    line 21
  2. Distinctive:

    Check what each kind of write does to the people who use the tracker, in this tracker, with this project's automations.

    line 46
  3. Distinctive:

    Before each write, read the card's updated time again.

    line 65
  4. Very specific:

    After each write, read the card back and compare it with what you wrote: the text, the author, and the updated time.

    line 72
  5. Distinctive:

    Keep a receipt for each write: the card, the text, the account that acted, the time, and the words of the grant that covered it.

    line 75

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Sync a tracker

You are done when the project can see what the card says and the tracker holds only what a grant covered: each card you will read again or whose requirements are tracked is a snapshot that the sources verbs can compare, and each write you made is read back and has a receipt. The tracker belongs to someone else, often a client, and its people see what you write. The engine holds no credential and calls no service: every read and write goes through a tool the person has already signed in to.

The project's own tickets are not this tracker. Link a card and a ticket; do not copy one into the other, so that neither holds a stale copy of the other.

Read a card

Read through the harness's connector, a command-line tool the person has signed in to, or an export the person gives you. Do not ask for a password or a token, and do not look for one.

A card read once to answer a question is read and answered. Where it will be read again or its requirements tracked, save it as a markdown snapshot, under one file name that stays the same for this card, with:

  • the card's key and its fields, each under its own heading;
  • the time the tracker says the card was last updated;
  • each comment under its own heading that holds the comment's id, with its author and time (the importer makes one item for each heading, so a comment kept under a shared heading would hide a new one);
  • the links to other cards;
  • the ids of the attachments, and their names.

Where a snapshot is kept, import it with outcomebound sources import --converted, under the same --name each time you read this card. Then a field that changed since the last read shows as a changed item, and a new comment as a new one, and gather-requirements accounts for each. A snapshot cannot show that it is current: it is the card as of the time you read it, and says so.

The snapshot holds the client's text. Keep it out of the repository unless the project says otherwise, and do not paste it into a public place.

Write to a card

Check what each kind of write does to the people who use the tracker, in this tracker, with this project's automations. A standing grant covers a write only when it names three things: the kind of write, the notifications and automations that write is known to start, and the conflict rule (what happens to a change someone else made since you read the card). A request for one named write in this session covers that write once you have shown the text. Any other write goes as a decision brief, with the old text, the new text, what it is known to start and what is UNVERIFIED.

  • A comment commonly notifies the card's watchers. A grant for comments in batches covers them when it says that they notify and who is told.
  • A description write commonly replaces the whole field, and a tracker may hold no lock, so a change made by someone else since you read it can be lost. A grant covers it only when it states the conflict rule, such as writing only if the updated time is unchanged.
  • A status transition can start the client's workflow automations, which may notify people, assign work, or close related cards. A grant covers it only when it names what the transition starts.

Authority to read is not authority to write, and a grant for one kind of write is not a grant for another.

Before each write, read the card's updated time again. If it moved since your snapshot, import a new snapshot and look at what changed before you write. That read cannot catch a change made after it: where the tool can make the write conditional on the version or updated time, do so; where it cannot, write only where the grant accepts that race, else give the person a draft. Unless a standing grant covers this kind of write, show the person the exact text first; under a batch grant, show the text once for the batch.

After each write, read the card back and compare it with what you wrote: the text, the author, and the updated time. A write whose read-back you did not make is UNVERIFIED.

Keep a receipt for each write: the card, the text, the account that acted, the time, and the words of the grant that covered it. The acting account is the person's own account unless they have given you another; say which.

Where the tracker answers that you are sending too fast, wait as long as it says, with one call in flight at a time, and read the card back before you send again: the first call may have succeeded. Where the read-back shows its effect, keep its receipt and do not send again; where it cannot tell, do not send again, and say why.

Where access is denied, or the harness has no tool that writes to the tracker, do not look for another way in. Write a draft file with the text and the card it belongs to, and give it to the person to paste.

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.