Borrowing it
Nothing to install: this file belongs to mazumba/opencode-dockerized. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/mazumba/opencode-dockerized/main/.opencode/config/skills/kanban-poller/SKILL.mdgit clone --depth 1 https://github.com/mazumba/opencode-dockerizedWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/mazumba/opencode-dockerized/kanban-poller)<a href="https://agentmods.dev/skills/mazumba/opencode-dockerized/kanban-poller"><img src="https://agentmods.dev/badge/skills/mazumba/opencode-dockerized/kanban-poller/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/mazumba/opencode-dockerized/kanban-poller"><img src="https://agentmods.dev/badge/skills/mazumba/opencode-dockerized/kanban-poller.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00057 | $0.07534 |
| Opus 5.5 | $0.00023 | $0.03014 |
| Sonnet 5.5 | $0.00011 | $0.01507 |
| Haiku 4.5 | $0.00006 | $0.00753 |
Grade A, and why
kanban-poller scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 175 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Kanban poller
Code: docker/plugins/kanban/poller/. Tests: docker/plugins/kanban/tests/. Commands it runs: docker/plugins/kanban/kanban.commands/.
Who writes what
- The poller makes every Linear write (state, labels, comments, reactions), all through the
--dry-runguard. Agents (ticket-worker|reviewer|investigator) havelinear_*denied. Agents write only code and the PR (worker), the GitHub review (reviewer), and the report file (investigator). - Before each work/fix/review/investigate run, and before the ticket is claimed or moved, the poller writes
<path>/.slim/worktrees/.tickets/<ID>.json(atomic; passed asctx.ticketFile; deleted when the job ends, also on failure; dry-run logs only). The snapshotstateis the state at write time, so a work run sees the pre-move state. Content: title, description, state, labels, project name, url, attachments, all comments oldest to newest (id,parentId,author,isPoller,createdAt,body), state history. Cleanup works per ticket worktree and never touches.tickets. - Every run ends with one last line
KANBAN-RESULT: <json>. The poller takes the LAST such line (tolerating surrounding backticks, bold, or quote markers) and rejects unknown fields or values:- Worker:
{"outcome":"review","pr":<n>}or{"outcome":"needs-human","step","reason"}.reviewis accepted only ifghfinds an OPEN PR on the branch with that number; thenAgent review. ElseNeeds human. - Reviewer:
{"outcome":"approve"}→Ready for merge;{"outcome":"changes","summary"}→ commentAgent review: changes requested (round n/2) — <PR>+ one-line summary (prefixedReview:if it starts withCI failedormerge conflicts), label,In Progress, orNeeds human("review limit reached") at round ≥ 2;needs-human→Needs humanwithAgent review: needs human — <step>: <reason>. - Investigator:
{"outcome":"done"}plus the report in<path>/.slim/worktrees/.tickets/<ID>.investigation.md; the poller postsAgent investigation:\n\n<report>(cut at 60k characters with a note) as a reply to the request and reacts ✅.needs-human, a missing report, or a bad marker = ❌ +Agent investigation: failed — <reason>. - Failed run (agent/server error, timeout, missing or invalid marker): one automatic retry, then
Needs human("agent result: marker missing|marker is not JSON|marker invalid", or the run failure). - Work/fix: retry any failure IF nothing was pushed (remote branch head before the run = after a fetch; unknown counts as pushed). Review: retry only an agent/server error, not a timeout or missing marker.
- Retry = put the ticket back to its start lane (work:
Ready for agent; fix: staysIn Progress+agent:changes-requested; review: staysAgent review), THEN commentAgent: retrying (<lane>) — <reason>. That comment is the persisted budget (max 1 per lane entry), counted only from comments authored by the poller (user.id= API user) at or after the entry (work: lastReady for agent; review: lastAgent review; fix: newest pollerchanges requestedcomment). The comparison is inclusive, so equal timestamps cannot grant another retry; a hard cap of 3 retry comments per lane per ticket ends loops whatever the timestamps. Logrun.retry. - A claim 👀 ends with its job (job
finally, every outcome, released even if removing the ticket files fails; a reaction that is already gone is ignored);staleClaimsignores tickets labelledagent:changes-requested. Claims are written first (claimIssue: add the new reaction, then delete older ones), then the move/relabel. - Session that could not be aborted (abort failed AND status still busy after 2 min):
Needs human("session could not be aborted"), never a retry or requeue. - Poller stop (aborted run): same put-back, no comment, no budget used, log
run.requeued; so a restart resumes the work. Investigations are not requeued (❌poller stopped).
- Worker:
- Marker extraction (
parseResultinselect.ts) is separate from the transport (runner.tsgives the stdout tail).opencode runprints the final reply text to stdout; if the format ever wraps or decorates the line, fix only the line pattern.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago Changed · +73 lines c310cacdd669
- 13d ago First seen · 102 lines · 57 tokens per session scan A aa81d8a23287
kanban-poller is a skill published in the GitHub repository mazumba/opencode-dockerized (6 stars, last pushed 2d ago), licensed MIT. It adds 57 tokens to every session and 7,534 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-13.
Other skills, from other repositories
bug-triage
Re-evaluate open bugs — priority vs severity, assign to sprints, surface systemic trends. Run when the count grows.
systemic-issue-triage
Trigger: new issue, bug report, triage, backlog, issue flood, community report, root cause, dead-end, blocked user. Attack issues by root class, never one-by-one; fixes must shrink the system, not grow it.
triage-issue-reports
Triage Slack issue reports with one thread-only verdict, evidence review, cause-aware routing, tracker dedupe, and fail-closed ticket creation. Use only from the configured Benny triage automation.
triage-issue
Intelligently triage bug reports and error messages by searching for duplicates in Jira and offering to create new issues or add comments to existing ones. When an agent needs to: (1) Triage a bug report or error message, (2) Check if an issue is a duplicate, (3) Find similar past issues, (4) Create a new bug ticket…
browse-flows
Browse Power Automate environments and flows interactively. Use when the user wants to browse, list, or explore their flows and environments.
operating-cadence
Designs the rhythm an organization runs on — which reviews happen weekly, monthly and quarterly, what each one decides, who owns the numbers presented, and how a signal at the front line reaches the people who can act on it. Use this to set up a management operating system, fix a meeting calendar that produces no…