agent-issues
/agent-issues
Write, improve and split GitHub issues, and write task briefs for subagents, so that coding agents can implement them without asking questions. Interviews for missing facts, verifies every pointer in the repo, states runnable done-checks, non-goals and stop-and-flag triggers, cuts oversized work into self-contained vertical slices, and runs a cold-read review when size or risk warrants it. Use when asked to write an issue or ticket, file work for an agent, turn context into issues, improve or split an existing issue, or brief a subagent. Triggers on "write an issue for", "file a ticket", "make
- Author
- smykla-skalski ↗
- Repository
- sai ↗
- Licence
- MIT
- Last repository change
- 27.10.2026
Assessment and security
- Files in this package
- 9
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 27.10.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Write, improve and split GitHub issues, and write task briefs for subagents, so that coding agents can implement them without asking questions. Interviews for missing facts, verifies every pointer in the repo, states runnable done-checks, non-goals and stop-and-flag triggers, cuts oversized work into self-contained vertical slices, and runs a cold-read review when size or risk warrants it. Use when asked to write an issue or ticket, file work for an agent, turn context into issues, improve or split an existing issue, or brief a subagent. Triggers on "write an issue for", "file a ticket", "make
Triggers from the description
Arguments
/agent-issues [description | issue url/number] [--draft] [--file] [--brief] [--cold-read | --no-cold-read]
Original author description
Write, improve and split GitHub issues, and write task briefs for subagents, so that coding agents can implement them without asking questions. Interviews for missing facts, verifies every pointer in the repo, states runnable done-checks, non-goals and stop-and-flag triggers, cuts oversized work into self-contained vertical slices, and runs a cold-read review when size or risk warrants it. Use when asked to write an issue or ticket, file work for an agent, turn context into issues, improve or split an existing issue, or brief a subagent. Triggers on "write an issue for", "file a ticket", "make
How it works
- Required guidance
- Arguments
- Core flow
- Execution contract
SKILL.md preview
--- name: agent-issues description: Write, improve and split GitHub issues, and write task briefs for subagents, so that coding agents can implement them without asking questions. Interviews for missing facts, verifies every pointer in the repo, states runnable done-checks, non-goals and stop-and-flag triggers, cuts oversized work into self-contained vertical slices, and runs a cold-read review when size or risk warrants it. Use when asked to write an issue or ticket, file work for an agent, turn context into issues, improve or split an existing issue, or brief a subagent. Triggers on "write an issue for", "file a ticket", "make this agent-ready", "split this issue", "break this down", "brief a subagent", or a pasted issue URL to tighten. Not for implementing an issue. license: MIT compatibility: Works in Claude Code, Codex, opencode and Copilot CLI. Needs the gh CLI to file or edit GitHub issues; without it the skill produces the draft only. A subagent tool is optional (the cold-read review falls back to an inline pass). argument-hint: "[description | issue url/number] [--draft] [--file] [--brief] [--cold-read | --no-cold-read]" allowed-tools: Agent AskUserQuestion Bash Glob Grep Read Write user-invocable: true metadata: short-description: Write issues that coding agents can finish --- # Agent issues An issue written for a coding agent is the agent's whole prompt. The implementer will not reliably ask what is missing, may start with no memory of the author, and will read a hint as an instruction. This skill produces issues and briefs that are complete, checkable and honest about what is verified. Four jobs share one rule set: - Create an issue from a description or pasted context. - Improve an existing issue. - Split an oversized issue or goal into self-contained slices. - Write a brief for a subagent (not filed anywhere). ## Required guidance Before taking any action, read [references/workflow.md](references/workflow.md) completely. It is the authoritative procedure and preserves every platform fallback, decision rule, template, command and output contract. Follow its sections in order and load the deeper references it names only at their stated gates. Paths in the workflow are relative to this skill directory. When a named tool, agent or interaction primitive is unavailable, use the workflow's compatibility fallback; never silently skip the behavior. ## Arguments Invocation arguments: `$ARGUMENTS` These are the description, issue URL or number and flags the user gave on invocation. The workflow reads them from here, not from its own text. If the line above is empty or still shows the literal placeholder, the client passed none: take the same values from the user's request. ## Core flow 1. Detect the mode and read the repository's conventions 2. Interview for missing facts, then research and verify every claim 3. Decide whether the work is one issue or several 4. Draft with the issue rules, or the repo's template 5. Run the self-check, then the cold-read gate if the threshold is met 6. Run the writing pass 7. Show the draft, get a go-ahead, file with native relationships 8. Read the result back ## Execution contract - Resolve the mode and flags before any side effect. - Execute the workflow sections in order. Load each reference at the step that names it, not before. - Never put an unverified fact in an issue as if it were verified. Mark it as a hypothesis or an open item. - Never file, edit or comment on GitHub before the user has seen the full draft and said go, unless the invocation carried `--file`. - Stop at every hard stop the workflow names and state the next action needed.
Files in this package (9)
- SKILL.md 4 KB
- agents/openai.yaml 1 KB
- references/cold-read.md 5 KB
- references/evidence.md 9 KB
- references/issue-rules.md 11 KB
- references/splitting-and-handoffs.md 8 KB
- references/templates-and-forms.md 4 KB
- references/workflow.md 10 KB
- references/writing-pass.md 2 KB
Versions
The database has no other skills named agent-issues.
Database snapshot: 13.11.2026 · package version 153cef27a8a0 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100writing-tickets
Write or rewrite an issue tracker ticket (Linear, GitHub Issues, Jira, or similar) that a person or coding agent will pick up. States the objective, evidence, decisions already made, hard rules, and done criteria, and leaves the how to the implementer. Use when the user says… - 100writing-session-briefs
Write a self-contained brief that hands a task to another session, subagent or teammate. Use when you are about to start a worker session, delegate to a subagent, split work across sessions, or leave a task for a future session to pick up. - 100dispatch
Hand ready GitHub issues to issue-worker subagents that implement, verify and open their own pull request, one worker per issue. Use when the user asks to dispatch, assign, or hand issues to agents, or to turn issues into PRs. - 90work-order-writer
Write a work order an agent or a contractor can execute end to end without asking a question. Use when delegating a task, briefing an agent, writing a spec, or when work keeps coming back wrong. - 100best-prompt-ever
Interviews a developer to turn a vague coding request into a complete, ticket-quality prompt that a coding agent can execute on the first try. It reads the repository first (read-only), asks short rounds of multiple-choice questions whose options come from the code, and ends… - 100agents-md
Write effective AGENTS.md files that give coding agents the context they need to work in a repository. Use when creating a new AGENTS.md, improving an existing one, setting up a repo for AI coding agents, or onboarding agents to a codebase. Triggers on: "write AGENTS.md",…