agentsclimarketplace

Issueflow

Skill jugol/issueflow-skills/skills/issueflow

Umbrella entrypoint for issue-driven Git Flow-lite work. Use to adopt issueflow, clarify a feature, continue an autonomous cycle, schedule parallel worktree lanes, govern a plan-first project, or route issue discovery, proof, merge, and compound learning.From its SKILL.md

Install
npx -y skills add jugol/issueflow-skills --skill issueflow

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.6 KB, as published. Nobody here has run it

issueflow

Route the request, preserve project truth, and keep implementation moving through issues to develop.

Hard gates

  • In autonomous cycles, keep the main thread as scheduler, reviewer, integrator, and merge-gate owner. Delegate implementation when workers are authorized; otherwise record a no-safe-delegation rationale before editing issue code.
  • issueflow parallel explicitly authorizes subagents and worktrees. Plain issueflow does not. For complex work, shape independent lanes first; request issueflow parallel only if missing worker authorization would force main-thread or serialized implementation.
  • Serialized controls dependency or merge order; it does not authorize main-thread implementation.
  • After dispatch, continue read-only scheduling instead of idling. Before wait_agent, scan and record Ready/Active, backlog/drafts, plan gaps, proof/test failures, review/feedback findings, solution follow-ups, and stale/deferred candidates. If nothing can dispatch, record one reason: no-candidate-after-minimum-scan, overlaps-active-lane, active-lane-budget-full, or blocked-on-worker-output.

Routing

  • New/adopting repo -> repo-bootstrap
  • Ambiguous feature, product, or UI request -> issue-brainstorm
  • Concrete problem or issue draft -> issue-raise
  • Readiness -> issue-intake
  • Branch/worktree/worker scheduling -> issue-dispatch
  • Web or Flutter implementation -> implement-web or implement-flutter
  • Proof -> qa-web-proof or qa-flutter-proof
  • issue/* -> develop -> merge-gate
  • Post-merge learning and cleanup -> issue-compound
  • develop -> main -> release-gate

Two tracks

  • Autonomous Cycle: on continue, iterate, resume, failures, reviews, or feedback, scan durable signals and create the next issue or multi-issue wave without waiting for the user to name every task.
  • Interactive Feature Intake: read repo context, clarify implementation-critical ambiguity, compare 2-3 approaches, obtain direction, then create an issue or wave. Concrete low-risk bugs may use a compact intake.

If work already changed before routing, backfill the issue, wave, and proof trail honestly.

Project truth and context

  • Read macro then micro context: current-state pointer, plan anchor, current wave/issue, recent proof, branch/worktrees, next action, then matching solution-index entries.
  • Treat the active plan as product truth. Classify work as aligned, extension, conflict, or deviation; update or record a decision before dispatching extensions/conflicts.
  • Plan work must name the target user goal experience and user-visible proof.
  • Prefer medium user-visible vertical slices and waves of independent lanes. Combine tiny related findings; do not create one-issue waves without a concrete reason.
  • Keep PLAN_ANCHOR.md and CURRENT_STATE.md bounded. Search archives before opening them; never bulk-load completed issues, old waves/proof, or all solution notes.
  • If automation needs user input, credentials, or a product/policy choice, pause it with a resume condition; do not delete it.

Finish contract

  • Start fresh issue work from develop; use issue/* branches/worktrees for implementation.
  • An issue finishes only after merge or PR/merge-queue handoff to develop, integrated proof as required, durable state updates, and returning the root checkout to develop.
  • Record follow-up work as issues or a concrete no-follow-up rationale.
  • Run issue-compound when work produced reusable learning, prevention rules, failed approaches, follow-up triggers, plan gaps, or stale active context.

References

Load only what the current decision needs:

  • Tracks/planning: ../../references/two-track-routing.md, ../../references/interactive-brainstorming.md, ../../references/plan-governance.md, ../../references/goal-experience-planning.md
  • Scheduling: ../../references/autonomous-wave-generation.md, ../../references/issue-sizing-and-scheduling.md, ../../references/parallel-delivery.md
  • Lifecycle/context: ../../references/branch-lifecycle.md, ../../references/automation-governance.md, ../../references/compound-learning.md, ../../references/context-governance.md, ../../references/history-compaction.md
  • Product shape: ../../references/vertical-slice-architecture.md, ../../references/experience-first-ui.md
  • Full policy conflicts/onboarding: ../../OPERATING-MODEL.md
  • Durable artifacts: ../../templates/

Keep looking

Skills are one crate of 608,666. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.