AI skills and plugins

Bundled AI Skills

imdone-cli ships the imdone plugin for Claude Code and Codex. The plugin carries imdone’s bundled skills, hooks, and progress-note helpers as one installable workflow bundle. imdone init, imdone clone, and imdone update install or update the plugin once per detected agent at user scope. An ancestor .imdone-cli.yml still selects backlog context, but it is not an agent-plugin installation directory.

Skill	Problem it solves
`imdone-ls`	You need to find the right local Jira or GitHub issues quickly without jumping back to the provider UI or hand-scanning the backlog.
`imdone-template`	A story exists but is missing the right template-backed structure, so you need a fast way to apply or repair the expected sections and attachment links on an existing issue.
`imdone`	You need the standard CLI-backed workflow for listing, reading, moving, templating, or creating Jira and GitHub issues, including reusable issue-creation defaults for coding agents.
`backlog-refinement`	TODO stories are too vague to build, so developers and product owners need a fast way to identify weak stories and tighten them into thin, executable slices.
`hypothesis-driven-development`	Teams start building before they agree on the value, success criteria, demo path, or execution plan, which leads to output without proof of outcome.
`ai-experiment-capture`	Teams try new AI coding practices but lose the problem, prompt, outcome, and evidence unless a developer explicitly captures the experiment as shared working memory, with no background monitoring.
`learning-loop-report`	It is hard to tell which in-progress stories are safe to hand off or how to brief others on the work without interrupting the team because planning context is missing or uneven across the work set.
`markdown-pdf-docker`	Teams have markdown reports and Mermaid diagrams, but no reliable way to turn them into polished PDFs without depending on host-specific tooling.
`cycle-time-report`	Teams want to inspect delivery speed from archived issue history and share the result as a markdown report, trend chart, and PDF without rebuilding the numbers by hand.

Imdone-bundled skills are available through the user-scoped imdone@imdone-bundled plugin. User-created or project-local top-level skills can still live under the configured backlog’s .claude/skills/ or .codex/skills/; when a supported agent is detected, imdone preserves those custom skill directories and links them into each configured ancestor without moving them into the bundled plugin.

Bundled AI Plugins

imdone-cli manages imdone@imdone-bundled once per detected Claude Code or Codex installation at user scope. It compares the installed and bundled semantic versions, then installs, updates, enables, or leaves an equal or newer enabled version unchanged. After user-scope verification succeeds, imdone removes only recognized legacy project registrations and managed plugin links; ambiguous paths and custom resources are preserved.

imdone update reports one result per plugin and agent with the scope, installed version, target version, decision, and cleanup status. JSON and YAML output preserve the same normalized fields for automation. If setup fails or a version cannot be verified, project cleanup stays blocked so the existing fallback remains available. Use the existing email or Discord links under Support & Feedback if the reported evidence does not explain the failure.

Plugin Problem it solves

imdone Developers start direct Codex or Claude sessions, disagree with AI output, add context, hit blockers, make decisions, complete meaningful work, or identify next steps during normal coding work, but that progress disappears unless the agent records a local progress note. The plugin ships the bundled imdone skills listed above, the imdone-ai skill for loading configured active-story context on demand, progress-note hooks, and target-selection prompts when no active story is configured.

Plugin	Problem it solves
`imdone`	Developers start direct Codex or Claude sessions, disagree with AI output, add context, hit blockers, make decisions, complete meaningful work, or identify next steps during normal coding work, but that progress disappears unless the agent records a local progress note. The plugin ships the bundled imdone skills listed above, the `imdone-ai` skill for loading configured active-story context on demand, progress-note hooks, and target-selection prompts when no active story is configured.

Direct agent sessions ask once at session start whether to enable event-triggered progress notes; the choice is local to that session. Sessions started with imdone ai or imdone hdd automatically enable progress-note detection and do not show the opt-in question. Declining or giving an unrecognized answer keeps detection off for that direct session, and a hook invocation without a stable session identity stays quiet instead of asking repeatedly.

For an enabled high-signal event, the hook compares current context with all locally available backlog stories and proposes the best-fit target plus up to three unnumbered candidates. Every displayed target includes its issue key and human-readable title or concise description, so contributors do not need to remember what a bare key means. Jira keys and numeric GitHub issues use the same local ranking path. Session and global stories are only ranking priors or no-match fallbacks; a stronger candidate can be proposed without replacing either state, and the contributor must still confirm or redirect before imdone note runs.