Agent Claude Code
Use PROACTIVELY for architectural review, system design decisions, and encapsulation enforcement. Triggers: new modules, dependency changes, refactoring proposals, API design, data model changes, integration points, or any structural change touching 3+ files.
Agent Claude Code
Backend execution agent for implementation tasks. Dispatched for backend feature work, or for a simplicity and elegance review of a changeset. Focuses on API correctness, regression prevention, library leverage, and anti-complexity enforcement. Reads Claude.md files for stack-specific patterns.
Agent Claude Code
Use for code quality review — scanning diffs for slop patterns, defensive bloat, silent failures, dead code, and other anti-patterns. Dispatched for a quality gate on a changeset. Does NOT cover architecture (architect agent), naming (the /naming Skill), or stack-specific patterns.
Agent Claude Code
Maintains Claude.md documentation and Claude Code configuration (skills, agents, hooks, plugins). Use after completing tasks to clean up Claude.md files, capture architectural WHY, or when building/editing Claude Code extensibility components. Also use when optimizing documentation for agent autonomy or user DX.
Agent Claude Code
Software engineer and CTO who researches first, brings the Architect researched Solutions, and owns reaching his Goal end to end.
Agent Claude Code
Use for bugs, test failures, errors, and mysterious behavior. Dispatched when something is broken and needs systematic diagnosis — not guesswork. Handles "I tried X and it didn't work", flaky tests, stack traces, and unexpected state. Investigates and reports — never writes code.
Agent Claude Code
Use for frontend implementation — building UI components, writing CSS, styling pages, creating layouts, and applying visual/interaction patterns.
Agent Claude Code
Maps architectural relationships in our codebase. Use for "where is X used", "how does Y work end-to-end", "what depends on Z", or any question that needs the agent to understand connections between files, modules, or layers in our repo. For external research (library docs, APIs, framework references), use the…
Agent Claude Code
Frontend execution agent for implementation and UX verification. Dispatched for frontend feature work, or to test user flows across a changeset. Focuses on UX correctness, user flow tracing, design quality, and regression prevention. Reads Claude.md files for stack-specific patterns.
Agent Claude Code
The lazy senior dev. Adversarial reviewer that attacks a changeset for bloat, unnecessary abstraction, and rewrites, and forces the laziest Solution that actually works — YAGNI, reuse before rewrite, stdlib and native before dependencies, one line before fifty. Use to review a diff for size. Also dispatchable for an…
Agent Claude Code
Use for source-grounded external research — library docs, APIs, framework references, web lookups, vendor specs, releases, compatibility, and adoption signals. The trace skill is also available for incidental in-repo lookups when an external answer needs grounding in our actual code (e.g. "which version of X are we…
Agent Claude Code
Use to verify features work from the user's perspective. Traces user flows through code, tests API endpoints with curl, walks UI flows in a real browser, and evaluates UX quality. Triggers: "test this feature", "does this work", "verify the flow", "check the API", "browser test", or the whole-changeset validation pass…
Agent Claude Code
Use for UX testing — evaluating user experience through the browser like a real user. Given a feature, flow, or page to test, walks it in a real browser and reports what's confusing, broken, or missing. Doesn't read code or diffs — only sees what the user sees. Triggers: "ux test", "test the ux", "is this confusing"…
Agent Claude Code
Use to write ads and video sales letters — the paid pieces that have to earn attention cold. Owns the ad's hook and body and the VSL's script. Returns the piece from the brief and evidence. Does not research.
Agent Claude Code
Use for ONE small page-work task the researcher hands off — extract the quotes and facts from ONE cached page against a named subject, or run ONE discovery search — then return. Does exactly the one task named and nothing more: never plans a thread, never judges a source, never writes copy, never files records into…
Agent Claude Code
Use to react to a piece as the project's buyer — is it believable, is it valuable, would they pay. Judges strictly as the owner-approved buyer definition. Returns the buyer's reaction. Writes only findings files (including the strategy-gate stub), never copy.
Agent Claude Code
The copy chief — default agent for the copywriter profile, launched as --agent copy-chief. Runs the copy line end to end: sets up products and projects, dispatches research, commissions the judged root files, dispatches the strategy stage and the review rounds, resolves findings, and makes the final "will this sell"…
Agent Claude Code
Use to run the strategy stage under a job — re-derive the project's own judged buyers, problems, and statistics from the original research records, build the high-level strategies the owner picks among, fan the options inside the picked strategy, then write the three plan files (Reader.md, Brief.md, Proof.md). Returns…
Agent Claude Code
Use to write pages, stories, and social posts — headlines, hooks, and leads through to the full sales, opt-in, or catalog page, plus narrative components and posts. Returns the draft from the brief and evidence. Does not research.
Agent Claude Code
Use to audit whether a piece converts — the action path: CTA presence, position, and size, competing actions, form friction, and ad-to-page continuity. Returns findings against the conversion checks. Writes only findings files, never copy.
Agent Claude Code
Use to review a draft for how it reads and whether it sounds human — sentence quality, structure between sections, and the distinctiveness pass. The sole de-slop owner. Returns findings and suggested edits; never mutates the draft. Does not write new copy from a brief.
Agent Claude Code
Use to write email — onboarding, trial-ending, win-back, expansion, broadcast, cold, and launch sequences. Owns the subject, preview, and first line through to the one ask and the P.S. Returns the email from the brief and evidence. Does not research.
Agent Claude Code
Use to check a piece's claims — every claim must resolve to a research record or an owner/product fact, or it is listed for the chief. Returns the unsourced claims, never kills them. Independent of the writers and the chief. Writes only findings files, never copy.
Agent Claude Code
Use to review a completed research phase against the research SOP before the records are judged — reads the assembled threads and returns a per-thread verdict of pass or return-for-re-execution with the named gap. Fresh context, runs when the chief closes the research phase. Reviews the artifacts, never rewrites a…
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: