Agent
Audits the persistent memory directory for decayed, superseded, or misplaced entries, archives the directory before any change, and deletes or rewrites what decay-checks confirm. Use when memory has accumulated stale referents, self-reversing entries, orphaned index lines, or live work items that belong in a tracker…
Agent
Inventories one shard of a prototype (routes/pages, data models, integrations/deps/stack, or a read-only bees map) with no judgement or classification — recording only. Spawned by pm-lead.
Agent
Parses harvest/assess requests, detects prototype shape, spawns discovery/separator/assessor/writer workers in sequence, and aggregates their reports without upgrading confidence tags. Spawned by /pm:harvest and /pm:assess.
Agent
Authors a structured PRD from the prd template, either interviewing the user with a questionnaire or grounding on a spec-harvest feature inventory. Spawned by /pm:prd and by pm-lead during /pm:harvest.
Agent
Walks the spec-harvest SDLC checklist (licensing, security, supportability) against a prototype, gathering evidence via read-only commands and tagging each row's confidence. Spawned by pm-lead.
Agent
Spot-verifies sampled inventory claims in code and classifies each item as a customer-value feature, a prototype shortcut, or both, drafting per-feature WHAT/WHY value statements. Spawned by pm-lead.
Agent
Composes the feature-inventory and SDLC-assessment markdown artifacts from the separator and assessor reports, self-validates against the source reports, and refuses to overwrite existing files. Spawned by pm-lead.
Agent
Processes beads tasks by polling bd ready, executing work, and syncing results. Use for automating task queues or AI-driven workflows.
Agent
Read-only collector for open Dependabot PRs. Use when listing and classifying open app/dependabot PRs before consolidation.
Agent
Cherry-picks Dependabot PRs into a consolidated branch, resolves keep-both conflicts, runs baseline-diff gates, and pushes a draft PR. Use when executing the mechanical consolidation after the collector has classified the PRs.
Agent
Read-only collector for open human-authored PRs. Use when listing and classifying a repo's non-bot pull requests and discovering its gate tasks before per-PR review.
Agent
Reviews one human-authored PR by running baseline-diff gates on main vs the branch and reviewing the diff against a rubric, then emits a structured verdict. Use when reviewing a single classified PR after the collector has run. Never edits code, never merges.
Agent
Questionnaire agent that interviews a user about a feature, writes a persona-aware Gherkin user story to docs/user-stories/ / .md, and self-validates the output against the qa parser. Spawned by /qa:new-story.
Agent
Validates Gherkin backend assertions against any non-Phoenix backend via HTTP, CLI, and log probing. Generic fallback when Tidewave is unavailable. Emits BEES REQUESTS blocks for deltas. Spawned by qa-lead.
Agent
TDD implementation worker. Receives a failing-test artifact from qa-test-writer and writes the smallest production code change to make it pass. Forbidden from modifying tests — that boundary is enforced by qa-lead diff inspection. Half of the adversarial fix pair.
Agent
Parses a Gherkin user story, detects the application stack, decomposes scenarios into worker tasks, aggregates worker reports, and routes deltas through bees-manager. Spawned by the /qa command.
Agent
Drives Gherkin UI scenarios against a running web app via Playwright MCP. Asserts Then clauses using the accessibility tree and emits BEES REQUESTS blocks for any deltas. Spawned by qa-lead.
Agent
TDD test-author worker. Reads a filed bees issue and writes ONE failing test that captures the bug. Forbidden from modifying production code — that boundary is enforced by qa-lead diff inspection. Half of the adversarial fix pair.
Agent
Validates Gherkin backend assertions against a running Phoenix application via Tidewave MCP. Asserts DB state, log lines, and runtime invariants. Emits BEES REQUESTS blocks for deltas. Spawned by qa-lead.
Agent
Uses Playwright MCP browser tools to extract brand tokens (colors, fonts, spacing) from websites and style guides.
Agent
Questionnaire agent that discovers audience, tone, purpose, and problem framing before building presentations.
Agent
Builds Slidev presentations consuming content strategy and brand config, applying presentation best practices and audience-appropriate templates.
Agent
Inspect WebAssembly binaries for validity, structure, imports, exports, WIT interfaces, and size.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: