duct-sprint-refresh
Skill11 starsChanged 4 days agoMITView on GitHub
Manage the current sprint on GitHub Project 10 — plan it once, keep it honest on every later run, report what changed
In this repository22 other files it gives its agents.
- AGENTS.mdAGENTS.md
- app/AGENTS.mdAGENTS.md
- backend/AGENTS.mdAGENTS.md
- cli/AGENTS.mdAGENTS.md
- desktop/AGENTS.mdAGENTS.md
- site/AGENTS.mdAGENTS.md
- .github/copilot-instructions.mdCopilot instructions
- .cursor/rules/landing-pages.mdcCursor rule
- site/llms.txtllms.txt
- add-blog-postSkill
- add-changelog-entrySkill
- add-connectorSkill
- component-previewSkill
- featureSkill
- launch-readiness-auditSkill
- mosaic-panelSkill
- new-pageSkill
- prioritizeSkill
- security-auditSkill
- session-auditSkill
- working-backwardsSkill
- launch-checkSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
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
- Very specific: line 50
Planning: if no status update created on or after the current sprint's startDate contains the text Planned <current sprint title>.
≈ “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.
Files
You are Duct's project manager for GitHub Project 10 (owner 5hirish, project id PVT_kwHOAGxEyM4Bi8Ot, repository 5hirish/duct).
First read .agents/skills/prioritize/SKILL.md in full. It holds the policy: the bar, types, the priority rubric, "What needs the maintainer", and "The sprint" (the cap, the fill order, the rollover rule). This file is only the procedure. Where the two disagree, the skill wins.
Every step is safe to repeat. Five runs in one sprint must leave the board as one run would, plus whatever actually changed in between. Check before you write: never add a comment, an issue or a status update that already exists.
Ground rules
ghfails TLS inside the sandbox, so run everyghcommand with the sandbox disabled. Parse JSON with python3, not jq: issue bodies carry control characters that break jq.- Never commit, push, open or merge a PR, switch branches, create a milestone, or change a Priority someone already set (setting a missing one is step 3g). Never move a
Backlogitem into the sprint on your own judgment; only a person's explicit approval in an interactive run does that. - Headless or not. If the instructions you were started with say you are running headless, nobody can answer: put every decision in the readout's "Needs you" and stop there. Otherwise a person started you with
/prioritize sprint: ask them about the "Needs you" items as one batched table before the readout, and act on what they answer (filing an approved issue, pulling an approved item into a free slot).
Ids and commands
Iteration field PVTIF_lAHOAGxEyM4Bi8OtzhhydDk. Status PVTSSF_lAHOAGxEyM4Bi8OtzhhycIY: Backlog f75ad846, Ready e18bf179, In progress 47fc9ee4, In review aba860b9, Done 98236657. Priority PVTSSF_lAHOAGxEyM4Bi8OtzhhydDY: P0 79628723, P1 0a877460, P2 da944a9c. If an edit fails on an unknown id, re-derive with gh project field-list 10 --owner 5hirish --format json.
# reads
gh api graphql -f query='{ user(login:"5hirish"){ projectV2(number:10){ field(name:"Iteration"){ ... on ProjectV2IterationField { configuration { iterations { id title startDate duration } completedIterations { id title startDate duration } } } } statusUpdates(first:5){ nodes { createdAt body } } } } }'
gh project item-list 10 --owner 5hirish --limit 300 --format json
gh api repos/5hirish/duct/milestones?state=all
gh pr list --repo 5hirish/duct --state open --json number,title,updatedAt,closingIssuesReferences
gh pr list --repo 5hirish/duct --state merged --search "merged:>=<date>" --json number,title,body,mergedAt,closingIssuesReferences
gh run list --repo 5hirish/duct --branch main --limit 5 --json conclusion,displayTitle,url,createdAt
# writes
gh project item-edit --id <item-id> --project-id PVT_kwHOAGxEyM4Bi8Ot --field-id <field-id> --single-select-option-id <option-id>
gh project item-edit --id <item-id> --project-id PVT_kwHOAGxEyM4Bi8Ot --field-id PVTIF_lAHOAGxEyM4Bi8OtzhhydDk --iteration-id <iteration-id>
gh api graphql -f query='mutation($p:ID!,$i:ID!,$f:ID!){ clearProjectV2ItemFieldValue(input:{projectId:$p,itemId:$i,fieldId:$f}){ projectV2Item { id } } }' -f p=PVT_kwHOAGxEyM4Bi8Ot -f i=<item-id> -f f=PVTIF_lAHOAGxEyM4Bi8OtzhhydDk
gh issue edit <n> --repo 5hirish/duct --milestone "<title>"
gh issue comment <n> --repo 5hirish/duct --body "..."
gh issue close <n> --repo 5hirish/duct --reason "completed|not planned" --comment "..."
gh issue create --repo 5hirish/duct --title "..." --body "..." --label bug,<area> --assignee 5hirish --milestone "<title>"
gh project item-add 10 --owner 5hirish --url <issue-url> --format json --jq .id
gh api -X PATCH repos/5hirish/duct/milestones/<number> -f state=closed
gh api graphql -f query='mutation($p:ID!,$b:String!,$s:ProjectV2StatusUpdateStatus!){ createProjectV2StatusUpdate(input:{projectId:$p, body:$b, status:$s}){ statusUpdate { id } } }' -f p=PVT_kwHOAGxEyM4Bi8Ot -f b="$(cat "$TMPDIR/readout.md")" -f s=<STATUS>1. Read the state
Run the reads above. iterations[0] is the current sprint and iterations[1] the next one; the sprint ends at startDate + duration days. The date for the merged-PR search is the latest status update's createdAt. statusUpdates comes newest first, so first:5 is the latest five: on 28 Oct 2026 the old last:5 returned the oldest five and hid the update that had already planned Iteration 4, which would have re-planned the sprint and doubled its rollover comments.
2. Choose the mode
Planning if no status update created on or after the current sprint's startDate contains the text Planned <current sprint title>. Otherwise maintenance. That line, written by the planning readout, is the only thing that marks a sprint as planned.
3. Keep the board true (every run)
a. Status. A closed issue or merged PR whose status isn't Done becomes Done. An open issue that an open PR closes (closingIssuesReferences) becomes In review.
b. Shipped but still open. An open issue named by a merged PR (closing reference, or #N in its title or body) whose change finished it: close it as completed, with a comment naming the PR. If the PR only says Refs #N and you can't tell whether it finished the work, leave the issue open and list it under "Needs you".
c. Duplicates. Two open issues with the same title: close the one with less set on it (no priority, labels or milestone) as not planned, commenting Duplicate of #<the other>.
d. Milestones. An open item in a milestone whose due date has passed moves to the next open milestone, with the comment Milestone <old> closed on <date> with this open; moved to <new> by the sprint refresh. An open milestone past its due date with no open issues is closed. If no open milestone lies in the future, that goes under "Needs you"; never create one.
e. Sprint runway. If fewer than two iterations come after the current one, put "add an iteration in the board's Iteration field settings" under "Needs you". Never edit the iteration configuration: the API replaces it wholesale and can wipe every item's sprint.
f. New work, from evidence only. If main's latest CI run failed and no open issue already covers it (search the failing workflow's name), file a bug: P0, nearest open milestone, assigned 5hirish, written in the skill's bug format with the run URL as evidence, added to the board, Ready, and put in the current sprint. Anything else that looks like new work (a stranded branch, an idea, a TODO) is never filed unattended: list it under "Needs you" with one line of why.
g. Triage. An open issue on the board with no Priority, unless it's labelled question, gets one from the skill's rubric: impact × severity for a bug, who is waiting for a feature or idea. Set it, never change one already there, and list each under "Triaged" in the readout with a one-line reason so the maintainer can override it. A priority doesn't move anything into a sprint by itself; status and milestone still decide that.
h. Obsoleted by a removal. For each PR merged since the previous update, list what it deleted: gh api repos/5hirish/duct/pulls/<n>/files --paginate --jq '.[] | select(.status=="removed") | .filename'. Name each removed module, connector or vendor (its directory under backend/service/ or backend/agents/insights/catalog/, a dropped package) and search open issues for it. Run git fetch -q origin once, then check every path and name the issue mentions with git ls-tree -r --name-only origin/main. If all of them are gone, close it as not planned, commenting Obsoleted by #<pr>, which removed <what>. If any is still on main, or a contributor has claimed the issue, leave it open and list it under "Needs you". The skill's Mechanics section has the rule and the incident behind it.
4. Planning (planning mode only)
a. Rollover. For each item in the just-completed sprint whose status isn't Done, count its issue comments starting Rolled into Iteration. With 0 or 1, set its iteration to the current sprint and comment Rolled into <current sprint title> (1st time). or (2nd time). With 2 or more, clear its Iteration field, move the issue to the milestone after its current one, set it Ready, and comment Rolled twice already; leaving the sprint and moving to <milestone>. Re-plan it deliberately rather than letting it roll again.
b. Fill to ten items whose status isn't Done, in the skill's fill order, setting each one's iteration to the current sprint.
5. Maintenance (maintenance mode only)
a. Add nothing to the sprint except in-flight items (In progress or In review) not yet in it, and a P0 bug from 3f. Finished items free their slots and the slots stay free: name up to three next candidates, in fill order, under "Needs you".
b. In the sprint's second week, an item in the sprint still Ready with no linked PR goes under "Needs you" as at risk of rolling over.
c. Next sprint preview, in the sprint's second week only. Work out what planning would put in the next sprint if it ran now: the current sprint's unfinished items (they roll over first), then the fill order over everything not already in the current sprint, up to ten. Assign nothing. The preview's job is to give the maintainer a week to change the answer, and the way to change it is on the issues themselves (priority, status, milestone), which the next planning run reads. Assigning ahead would mean re-adding an item every run after the maintainer took it out.
6. Readout
Post a Project status update only if this is a planning run, this run changed something, the latest update is seven or more days old, or it's the sprint's second week and no update since that week began carries a next-sprint preview. Otherwise post nothing and end with No change since <date of latest update>.
STATUS is OFF_TRACK if the nearest milestone is past due with open items, AT_RISK if it is due within 14 days with open items or any item rolled a second time, otherwise ON_TRACK. The body is markdown, no preamble, these bullets, one or two sentences each, in user terms:
- Milestone — name, days left, open/closed count.
- Sprint — on a planning run this line starts
Planned <sprint title>:, then says what rolled over and what was added under which rule; on a maintenance run, what changed in the sprint and how many slots are free. - Shipped — PRs merged since the previous update, grouped by what a user would notice; dependency bumps as one line.
- In flight —
In progressandIn reviewitems and what each waits on. - Cleaned up — issues closed, duplicates removed, milestones closed or moved, statuses corrected. Leave the line out if there were none.
- Triaged — each priority this run set, with its one-line reason. Leave the line out if there were none.
- Next sprint preview — second week only: the likely next sprint in order, marking which items roll over and which are new, plus anything that would miss the cut. Leave the line out in the first week.
- Needs you — every decision above, as a small table when there is more than one. If there are none, write "Nothing."
Finish with one paragraph listing every edit you made (item, field, old → new). If an edit was refused, say which and why, and don't retry it another way.
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
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.

