Build and maintain a retrieval-grounded knowledge agent from product documentation: design the corpus, split oversized sources into retrievable parts, scan for redaction and voice problems, prove after upload that the agent can actually read every part, and run an evaluation set before sharing it.
Plugin rag-agent-builder from marketplace pm-agent-toolkit in ryanhoggpm/pm-agent-toolkit. Commands read from the repo's manifest — 1 month ago.
rag-agent-builderBuild and maintain a retrieval-grounded knowledge agent from product documentation: design the corpus, split oversized sources into retrievable parts, scan for redaction and voice problems, prove after upload that the agent can actually read every part, and run an evaluation set before sharing it. U
Review an OpenAPI spec for structural defects (wrong HTTP verbs, empty responses, broken security schemes), CRUD and documentation gaps, and coverage of stated product requirements, producing a severity-tagged report with a prioritized fix list.
1 skill
Set up and run multi-session frontend development in Bolt.new against a real backend API, producing a Bolt Project Knowledge document, a PROGRESS.md session tracker, a Vite proxy config for CORS-free live API calls, and a Bolt-to-local-review production handoff.
1 skill
Search prior work across the workspace's outputs/, context/, and memory folders before starting any task that might duplicate it, returning up to 15 ranked results with file paths, modified dates, and matching snippets.
1 skill
Designs and writes well-scoped working agents, either Claude Code subagents (.claude/agents/*.md) or Claude Managed Agents (API config), from a structured design interview covering scope-as-refusals, routing description, least-privilege tools and permission policies, output contract,...
1 skill
Build a new Claude skill to a 10-rule authoring standard (routing-engineered description, read-first table, shipped output template, exit checklist), or audit an existing skill against that standard.
1 skill
Run a marketing draft through a simulated focus group of weighted buyer personas, producing per-persona reactions (what each would cut or share), a conflict-resolving synthesis, a shipped-claims check, and a revised draft calibrated to audience weighting.
1 skill
/build: one unit of work from a request to a merged PR through BRIEF, BUILD, CLOSE and SHIP, with its fourteen agents, runtime scripts and the subagent no-push guard.
14 agents · hooks
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.