From utils
Candango Executor. Execute approved feature issues autonomously through a main-agent orchestration loop. Use when the user wants the agent to identify executable issues, generate an HTML dashboard with dependency tree, Kanban status, and execution history, delegate each issue to a TDD executor subagent, delegate validation to an independent read-only validator subagent, loop until validation passes, run final repo validation, and communicate all user-facing updates and reports using caveman style.
How this skill is triggered — by the user, by Claude, or both
Slash command
/utils:candango-executorThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Main agent orchestrates. Executor subagent edits. Validator subagent reviews only.
Main agent orchestrates. Executor subagent edits. Validator subagent reviews only.
$tdd and may edit files.done, regardless of its previous status, main agent rechecks every backlog issue and starts any issue that is now unblocked.$caveman style: terse, high-signal, no fluff.$candango-discover.Read:
docs/agents/autonomous-feature-execution.mddocs/agents/feature-issue-tracker.mddocs/agents/feature-validation.mddocs/agents/feature-docs.mdIf final validation command is missing, discover from CI first. If still unknown, ask:
Is there any command you want me to run at the end of feature development?
Select issues that are:
backlog or explicitly selected by userBuild dependency graph. Detect cycles before starting.
Use scripts/feature_dashboard.py.
Create:
docs/features/<feature-slug>/state.jsondocs/features/<feature-slug>/index.htmlplan.md, uat.md, issues.md, and design/.Dashboard must show:
backlog, in-progress, validation, uat, doneSee references/dashboard-state.md.
Run continuously while selected issues remain in backlog, in-progress, or validation.
At startup, and after any issue transitions to done, rescan all backlog issues:
backlog, and all blocked_by issues are complete enough for its dependency rule.validation, uat, or blocked states; only done releases new starts after the initial scan.For each startable issue:
in-progress.references/executor-prompt.md.validation.references/validator-prompt.md.uat when UAT/manual acceptance remains, else done.in-progress and send validator report to executor.done, immediately rescan all backlog issues and start any newly unblocked issue.If UAT is manual, hand off to $candango-uat-runner before done.
After all selected issues are done:
Pause and ask user when:
Guides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub lucasbayma/skills --plugin candango-skills