prd-please

Capture implementation learnings during build phases and propose amendments to the Strategic PRD (upstream) and/or build documentation (lateral). Use this skill after completing a build phase, encountering repeated implementation issues, resolving non-trivial build problems, or at the end of a build sprint. Trigger when the user says things like "capture build learnings", "log what happened during the build", "this phase was harder than expected", "update docs based on what we learned", or "build retrospective". This skill closes the feedback loop between implementation and specification.

Generate a sub-agent invocation brief from the standard template before spawning any impl, validator, or learner sub-agent in the build loop. The brief bakes in runaway mitigations, file-ownership contracts, role-typed permissions, token budgets, and a structured report format. Use this skill whenever the orchestrator is about to spawn a sub-agent, or when `prd-validator` or `prd-learner` is preparing an agent invocation. Trigger when the user says things like "brief a sub-agent", "spawn an impl agent", "fan out", or "draft a brief". The orchestrator never writes prose briefs by hand once this skill exists; it fills slots.

Generate or regenerate a project's curated build context pack — the single `CONTEXT-PACK.md` that sub-agents read instead of the full PRD + ADR + docs corpus. Use this skill at the start of an orchestration cycle, after a PRD or ADR amendment is applied, or whenever the pack predates the latest contract change. Trigger when the user says things like "regenerate the context pack", "refresh context", "build the context pack", or "the pack is stale". The pack is a derived artifact: every line comes from a tracked file, and regenerating it changes freshness, never meaning.

Emit a single, scannable, ranked decision package from a project's `LOOP-STATE.md` decision queue, replacing one-at-a-time interruptions. Use this skill when 3+ decisions are queued, when learner/apply debt is near the ceiling, when the next batches all depend on the same open decision, or at the end of a session with open decisions. Trigger when the user says things like "decisions", "what do you need from me", "batch the decisions", or "decision package". Humans review batched, ranked decisions far better than a steady stream of interruptions — the orchestrator should forecast the bottleneck and emit the package before velocity hits the wall.

Capture implementation learnings and codify them back into the Strategic PRD so future agents and team members inherit the knowledge. Use this skill whenever you need to close the feedback loop after building — after a validation failure, a bug is resolved, repeated mistakes in a session, or at the end of a sprint as a retrospective step. Trigger when the user says things like "capture learnings", "update the PRD based on what happened", "log what went wrong", "retrospective", "we keep hitting this issue", or "add this to the PRD". Always use this skill when the goal is turning implementation experience into permanent PRD rules.

Read, write, and update a project's orchestration loop state at `<project_scope>/LOOP-STATE.md` — current batch, decision queue, fan-out registry, debt counters, and workaround signatures. Use this skill at the start of any orchestrated build cycle, whenever another execution-loop skill needs the state file, or when the user asks where the build stands. Trigger when the user says things like "loop state", "where are we in the loop", "update loop state", "log a decision", "log a workaround", or "initialize the build loop". The loop state externalizes the impl → validate → learn → apply → commit cycle so it survives session boundaries.

Derive executable tasks and client-facing views from a Strategic PRD. Use this skill whenever you need to break a PRD down into buildable work items, generate the initial task backlog for a new engagement, map an incoming client request against the existing PRD, or regenerate task contexts after a PRD amendment. Trigger when the user says things like "generate tasks from the PRD", "what should we build first", "break this into tasks", "a client request just came in", "the PRD changed — update the tasks", or "start the sprint". Always use this skill when the goal is turning a Strategic PRD into actionable, scoped work items.

Translate a completed Strategic PRD into Al Dente documentation templates for AI-native SaaS building. Use this skill when a PRD is approved and ready to move into the build phase using the Al Dente boilerplate. Trigger when the user says things like "translate to Al Dente", "generate build docs", "prepare for build", "create Al Dente docs from PRD", "populate Al Dente templates", or "bridge PRD to implementation". This skill produces the 7 documentation files that Al Dente's build phases consume.

Validate that implementation work satisfies the requirements, acceptance criteria, and guardrails defined in the Strategic PRD. Use this skill whenever you need to check compliance after a task or batch is implemented, during code review, or as a pre-merge gate. Trigger when the user says things like "validate this against the PRD", "check if this meets the requirements", "run a compliance check", "is this ready to ship", "does this satisfy the acceptance criteria", or "pre-merge check". The validator is adversarial by design: it executes typechecks, builds, and tests rather than trusting self-reports, and treats a dishonest fallback as a hard failure. It never builds or modifies code — it only checks.

Create and edit Strategic PRDs from human-provided context. Use this skill whenever you need to start a new Strategic PRD from scratch, revise an existing one due to scope changes, incorporate amendments proposed by the prd-learner skill, or make any structural changes to a PRD document. Trigger when the user says things like "write a PRD", "create a product spec", "update the PRD", "the scope changed", "incorporate these learnings", or "draft requirements for X". This is the only skill authorised to write or modify Strategic PRD documents — always use it when a PRD is the output.

Walk a new user through setting up prd-please for their company. Detects existing configuration, supports either an in-repo COMPANY.md or a pointer to an external company context file (e.g., an internal repo or vault), populates the schema section by section with teach-through-choice questions, and wires CLAUDE.md so downstream skills can find the profile. Refers users to sibling please-family tools when a question is upstream of prd-please's scope. Trigger this skill when the user says things like "I just cloned prd-please, where do I start", "set this up for my company", "onboard me", "help me get started", "I want to use prd-please but I don't have a company profile yet", or when a fresh checkout has no `CLAUDE.md` and no populated `companies/<slug>/COMPANY.md`.

Run a structured pre-authoring interview to uncover what is known, what can be inferred, and what is genuinely unknown before writing a Strategic PRD. Produces a Discovery Brief that prd-author ingests as its primary input. Use this skill before prd-author on any new engagement or when the input material is thin, one-sided, or missing key domain perspectives. Trigger when the user says things like "I want to write a PRD but I'm not sure where to start", "here's a rough idea", "I have a transcript but no context on the commercial side", "let's figure out what we know before writing", or "run discovery on this". Skip this skill if the human already has rich, multi-domain source material and is ready to go straight to authoring.