ddaanet
@ddaanet
Public Skills
orchestrate
by ddaanet
Execute prepared runbooks with plan-specific agents and mechanical verification gates
runbook
by ddaanet
Decompose a design into executable implementation steps. Triggers on /runbook or when a design needs step-by-step planning. Creates runbooks with per-phase typing (TDD cycles, general steps, or inline pass-through) for weak orchestrator execution.
design
by ddaanet
Triage complexity and design implementation approach. Triggers on /design, architecture or implementation planning requests, or tasks needing complexity assessment. Triages simple/moderate/complex, produces design documents for complex jobs, routes moderate to /runbook.
worktree
by ddaanet
Manage git worktrees for parallel task execution. Triggers on "create a worktree", "set up parallel work", "merge a worktree", "branch off a task", or uses the wt, wt merge, or wt-rm shortcuts. Worktree lifecycle: creation, focused sessions, merge ceremony, cleanup, parallel task setup.
inline
by ddaanet
Sequence inline execution lifecycle: pre-work, execute, post-work. Triggers on /inline, "execute inline", "run task", or when /design and /runbook route Tier 1/2 execution-ready work. Wraps corrector dispatch, triage feedback, and deliverable-review chaining.
handoff
by ddaanet
Save session state for the next agent. Triggers on "handoff", "h", "hc", session update, or agent switch requests. Writes session.md with completed tasks, pending work, blockers, and learnings. Not for Haiku orchestrators — use /handoff-haiku instead.
commit
by ddaanet
Create git commits for completed work with short, dense, structured messages. Use --context flag when you already know what changed from conversation.
prioritize
by ddaanet
Score and rank pending tasks by priority. Triggers on "prioritize", "score the backlog", "rank tasks", "reprioritize", or task ordering needs. WSJF-adapted scoring producing priority-ordered tables with parallel batches.
handoff-haiku
by ddaanet
Internal skill for Haiku model orchestrators only. Not for Sonnet or Opus — use /handoff instead. Mechanical session context preservation without learnings judgment.
prime
by ddaanet
Load plan artifacts and recall for ad-hoc plan work outside workflow skills. Triggers on "/prime", "load plan context", "prime the plan", or when working on plan artifacts without a workflow skill (design, runbook, orchestrate). Reads existing plan files, then chain-calls /recall.
deliverable-review
by ddaanet
Review production artifacts after plan execution. Triggers on "review deliverables", "deliverable review", "artifact review", or after completing plan execution. Two-layer review (delegated per-file + interactive cross-project) producing severity-classified report grounded in ISO 25010 / IEEE 1012.
memory-index
by ddaanet
Memory recall via Bash transport for sub-agents without Skill tool access. Provides when/how entry index and resolution examples.
requirements
by ddaanet
Capture and document requirements for implementation. Triggers on "capture requirements", "document requirements", "what do I want to build", or feature discussions without clear documentation. Produces requirements.md artifact for design and planning phases.
how
by ddaanet
Recall procedural knowledge from project decisions. Triggers on "how to do X", step-by-step techniques, or implementation guidance needs. Invoke with "/how <trigger>".
when
by ddaanet
Recall behavioral knowledge from project decisions. Triggers on "when to do X", situational patterns, or decision content for recognized situations. Invoke with "/when <trigger>".
recall
by ddaanet
Load project decisions into session context before working. Triggers on "recall", "load context", "load decisions", or when starting work on a topic with relevant decisions. Cumulative across invocations, tail-recursive within each.
reflect
by ddaanet
Diagnose agent deviations and rule violations via root cause analysis. Triggers on "reflect", "diagnose deviation", "root cause", "why did you do X", "what went wrong", or "RCA". Must run in the session where the deviation occurred.
review-plan
by ddaanet
Review runbook quality for TDD discipline, step clarity, and LLM failure modes. Detects prescriptive code in GREEN phases, validates RED/GREEN sequencing, checks prerequisite validation, script evaluation, and step clarity. Use when reviewing runbooks after generation by /plan. Use when the user asks to "review runbook", "check runbook for prescriptive code", "validate RED/GREEN discipline", "check for implementation anti-patterns", "check LLM failure modes", or when /plan Phase 1/3 delegates to runbook-corrector agent.
ground
by ddaanet
Ground claims with external research before asserting methodology. Triggers on "ground", "create a scoring system", "design a methodology", "build a framework", or "synthesize best practices". Parallel diverge-converge research preventing confabulated structures.
codify
by ddaanet
Consolidate learnings into permanent documentation. Triggers on "codify", "consolidate learnings", or when learnings.md approaches its soft limit. Transforms session learnings into CLAUDE.md fragments, decision files, and skill references.
next
by ddaanet
Find pending work in secondary locations. Triggers on "what's next?", "any pending work?", or when no pending work exists in already-loaded context. Checks shelf, todo.md, ROADMAP.md.
doc-writing
by ddaanet
Write or rewrite project documentation. Triggers on "write a README", "write documentation", "document this project", "docs are stale", or documentation creation requests. Explore, draft, reader-test with fresh agent, fix gaps, review.
token-efficient-bash
by ddaanet
Write compact bash scripts using exec tracing pattern. Triggers when writing bash scripts with 3+ sequential commands. The exec 2>&1 + set -xeuo pipefail pattern eliminates echo statements via automatic command tracing, reducing script size by 40-60%.
project-conventions
by ddaanet
Project prose and output conventions for sub-agents. Injected via skills: frontmatter — not user-invocable. Bundles deslop (prose quality), token economy, and tmp-directory rules.
opus-design-question
by ddaanet
Consult Opus for design and architectural decisions. REQUIRED instead of AskUserQuestion. Triggers on approach choices (A vs B), trade-off analysis, pattern selection, or structural decisions — use this skill to consult Opus instead of asking the user. Only ask the user for subjective preferences, business logic, scope changes, or unclear requirements.
error-handling
by ddaanet
Bash error handling rules and token-efficient exception patterns. Loaded when agents need error handling guidance for bash command execution.
gitmoji
by ddaanet
Select gitmoji emoji for commit messages. Triggers on "use gitmoji", "add gitmoji", "select gitmoji", or when creating commit messages requiring gitmoji prefixes. Semantic matching of commit intent to emoji.
plugin-dev-validation
by ddaanet
Domain validation criteria for plugin components. Consumed by corrector during artifact review.
release-prep
by ddaanet
Validates git state, runs quality checks, updates documentation, and assesses release readiness before the human executes the release command. Invoked when the user asks to "prepare release", "release prep", "ready to release", "pre-release check", "check release readiness", or mentions releasing a package.
shelve
by ddaanet
Archive session context to todo list and reset for new work
review
by ddaanet
Review in-progress changes for quality and correctness
brief
by ddaanet
Transfer context (scope changes, decisions, conclusions) to a worktree task via plans/<plan>/brief.md. Triggers: "brief a worktree", "write a brief", "/brief <slug>", "transfer context to worktree".