desplega-ai

implementing

Plan implementation skill. Executes approved technical plans phase by phase with verification checkpoints.

phase-running

Execute individual plan phases as background sub-agents for context-efficient implementation.

Data & Documents Listed

Execute a single DAG step as an autonomous background sub-agent. Sibling of phase-running for DAG plans produced by v-planning. Reads a step-<n>.md file directly, atomically claims it via frontmatter status, runs the three-bucket Success Criteria, and reports back. Spawned by v-implementing or by /run-step.

Data & Documents Listed

v-implementing

Parallel DAG-plan implementation skill. Reads a v-planning plan directory (root.md + step-<n>.md files), topologically schedules ready steps, and fans them out as parallel sub-agents. Use whenever the user invokes /v-implement, points at a plan directory produced by /v-plan, or asks to "run the parallel plan", "implement the DAG", or "fan out the steps" — even without those exact words. For linear plans (single .md file), use `implementing` instead.

ask-user

Conventions for asking the user questions during a workflow via the AskUserQuestion tool. Other skills reference this to keep prompts consistent and structured.

brainstorming

Interactive exploration of ideas through Socratic Q&A. Produces progressive documents that serve as lightweight pre-PRDs feeding into research.

Code & Development Listed

code-reviewing

Two-axis code review of a diff — Standards (is the code well built?) and Spec (does it actually solve the stated problem?) — run as parallel sub-agents and reported separately, never merged. Use when reviewing implementation output (a phase, a wave, a branch, a PR), when the user says "review the code", "review the diff", "review this phase", or when another desplega skill (implementing, v-implementing, verifying, delegate-work) calls for a per-phase or final code review. For reviewing DOCUMENTS (plans, research, brainstorms), use desplega:reviewing instead.

comms

Re-express something so it lands — re-explain your last reply in plain casual language, show the current topic visually (diagram, code-shape sketch, HTML artifact), do both at once, or rewrite artifact text into unambiguous ASD-STE100 English. Use when the user runs /desplega:comms, types /bro, says 'say it simpler', 'bro what', 'I don't follow', 'show me', 'draw this', 'visualize', 'disambiguate this', or 'STE100 rewrite'. Not for creative or marketing copy.

delegate-work

Executor routing policy for ALL delegated work — pick the right model for every sub-agent and the right Codex variant for every implementation slice. Use whenever you are about to spawn a sub-agent/Task (research, review, QA, UI, search), whenever a desplega skill (implement-plan, v-implement, run-phase, run-step, research) is choosing executors for phases/steps, or when the user says "implement with codex", "delegate this", "which model should do this", or similar. Supersedes codex-implement (its worktree/exec mechanics live here).

Code & Development Listed

design-docs

Create and maintain living per-system design docs at thoughts/*/design-docs/<system-slug>.md — Purpose, Glossary, numbered testable Invariants (normative for the code-reviewing Spec axis), Boundaries & Non-goals, Interfaces, Decision log, Amendment log. Use when the user says "create a design doc", "write down the invariants", "document the design/contract of X", or when researching/planning/one-shot finds a design doc for the touched system and needs the read-and-abide rules.

engineering-standards

Code-shape policy for ALL code that gets written or reviewed — the senior-engineer bar for minimal complexity. Use whenever you are about to write code, design a module or abstraction, draft a plan phase that shapes code, or review a diff; also when the user asks "is this over-engineered", "can this be simpler", or pushes back on complexity. Sibling of desplega:delegate-work (which picks WHO executes; this skill defines WHAT good output looks like).

improve-agents-md

Improve (or bootstrap) an AGENTS.md / CLAUDE.md file using `<important if>` conditional blocks so the agent actually attends to the right guidance at the right time. Use this skill whenever the user mentions AGENTS.md, CLAUDE.md, agent instructions, project rules for AI, "my claude config", onboarding docs for agents, or asks to tighten / shorten / audit / rewrite an existing one — even if they don't explicitly say the filename. Also use when the user complains that an agent keeps ignoring their project rules.

learning

Compounding knowledge across projects and teams. Captures, searches, and promotes institutional learnings via tiered backends (local/qmd/agent-fs).

one-shot

Small-scope plan+implement in one session. Assess the prompt, optionally one bundled round of questions, maintain a lightweight yolo plan while implementing, verify, quick code review, commit. Use when the user invokes /one-shot, or asks to "just build/fix/add" something small without the full create-plan → implement-plan pipeline. Small means ≤ ~3 phases inside one subsystem — anything bigger escalates to /desplega:create-plan.

planning

Implementation planning skill. Creates detailed technical plans through interactive research and iteration.

Testing & QA Listed

qa

Functional validation skill. Captures test evidence (screenshots, recordings, links) and produces QA reports in thoughts/*/qa/.

questioning

One-shot question answering using the research process. Answers inline without generating documents, then offers handoff to brainstorm or research.

researching

Comprehensive codebase research skill. Documents codebase as-is by spawning parallel sub-agents and synthesizing findings into research documents.

Code & Development Listed

reviewing

Structured critique of research, plan, and brainstorm documents for completeness, gaps, and quality.

tackle-gh-comments

Work through every review comment on a GitHub PR — fetch the threads, verify each claim against the code, fix what is real, then reply and mark each thread resolved before pushing. Use when the user says "tackle the PR comments", "address the review", "handle the bot comments", "check the comments on the PR", "resolve the review threads", or invokes /tackle-gh-comments. Project-agnostic; works in any repo with a GitHub remote.

Testing & QA Listed

tdd-planning

TDD-focused implementation planning. Creates plans with strict Red-Green-Commit/Rollback cycles for each step.

v-planning

Vertical / parallel implementation planning skill. Creates DAG-structured plan directories where each step is an independent, QA-able vertical slice that sub-agents can pick up and implement in parallel. Use whenever the user wants a plan that fans out (multiple independent features), invokes /v-plan, or asks for a "parallel plan", "DAG plan", "vertical plan", or "plan that can be parallelized" — even if they don't say those exact words. Prefer the linear `planning` skill for strictly sequential work.

verifying

Post-implementation plan verification. Cross-references plans against actual changes for completeness and accuracy.

brain-expert

Personal knowledge management expert for the brain CLI. Use when users want to capture notes, search their knowledge base, or manage their second brain.

Data & Documents Listed

file-review

File review tool — launch GUI, process comments, or install. Use when user mentions file-review, reviewing files, leaving comments, or processing review comments.

API & Backend Listed

oapi-expert

OpenAPI CLI expert for @desplega.ai/oapi. Use when users want to register OpenAPI specs, execute API requests, manage auth profiles, explore API endpoints, or work with REST APIs from the terminal.

feedback

Give Radically Candid feedback to the user — kind, clear, specific, and sincere — based on Kim Scott's framework (Care Personally × Challenge Directly). Use this whenever the user asks for your opinion, review, critique, or pushback on a plan, idea, code, writing, design, or decision; when the user invites honest disagreement ("tell me what you really think", "be honest", "poke holes in this", "am I missing something"); when something the user just said or did looks wrong and silence would be Ruinous Empathy; or when the user explicitly invokes /radical-candor:feedback. Prefer this skill over generic agreeable responses any time feedback would be more useful than validation. Default to a short headline reply and offer to expand — don't dump the full analysis unprompted.

remarkable-expert

reMarkable tablet expert. Use when users want to list, download, or upload files to their reMarkable tablet.

swarm-expert

Multi-agent coordination expert for agent-swarm MCP. Use when the user asks about swarm coordination, delegating tasks to agents, checking swarm status, agent messaging, or managing multi-agent workflows.

wts-expert

Git worktree management expert for @desplega.ai/wts. Use when the user asks about git worktrees, wts commands, worktree workflows, or wants help managing multiple branches simultaneously.