Agents, Commands and Pipelines for Claude code & other harnesses
Agent Claude Code
Generates a coherent alternative frame for the problem an artifact solves — one that shares none of the artifact's assumptions. Not a competing solution but a competing understanding of the problem itself. The most conceptually aggressive agent in the perspective family. Decision - FRAMEAWARE/FRAMELOCKED.
Agent Claude Code
Finds terms and phrases used with multiple meanings within the same artifact — not vagueness (imprecision), but genuine polysemy where the same word carries different meanings in different sections. Destroys handoffs. Corrupts scoring. Produces inconsistent behavior in automated systems that parse the artifact…
Agent Claude Code
Reads the artifact from the position of someone afraid of its failure modes — someone whose career depends on it not failing. Surfaces concerns that hide beneath confident language - unhandled edge cases, silent failure modes, undisclosed dependencies, untested assumptions. Labels findings by anxiety register …
Agent Claude Code
Validates API contract consistency between documentation, types, and implementation. Catches contract drift, breaking changes, and documentation staleness. Required for APIs consumed by external clients or other services. Prevents integration failures.
Agent Claude Code
Performs Aristotelian four-cause decomposition on any artifact — code, specs, plans, architectures, or documents. Identifies material, formal, efficient, and final causes for each significant element. Distinguishes essential from accidental properties. Assesses whether the artifact's telos is coherent and its means…
Agent Claude Code
Surfaces implicit assumptions buried in any artifact — agent definitions, prompts, business plans, technical specs, workflows, or documents. Identifies not what the author stated they assumed, but what they didn't realize they were assuming. Produces a ranked assumption inventory with fragility scores. Decision …
Agent Claude Code
Consume the finding sets of upstream security agents and compose them into end-to-end attack chains, assessing composed severity, required attacker capability, and the detection and containment gaps along each path. Re-ranks individually-minor findings by the severity they reach when chained, produces a prioritized…
Agent Claude Code
Models the perspective of someone who must use the artifact and has no exit option. Surfaces friction that voluntary user research structurally misses because the people experiencing the worst friction aren't the ones giving feedback — they can't leave. Identifies captive contexts, maps unreported abrasions, traces…
Agent Claude Code
Traces execution and communication chains end-to-end through codebases. Given an entry point (concept, endpoint, function, event), follows every hop, transformation, and handoff to produce a complete chain map with boundary crossings marked. Decision - TRACED/FRAGMENTED.
Agent Claude Code
Diagnoses why a system survives failure. Performs causal attribution of observed resilience behavior — recovery, degradation, absorption — to the mechanism that actually carries it, and renders per-defense operating status - working-as-designed, compensated-for, inert, or unproven. Acquires evidence by a three-rung…
Agent Claude Code
Surveys the failure surface of a system and produces a prioritized chaos-experiment agenda. Partitions the system into failure domains, classifies each by rehearsal depth — REHEARSED, DEFENDED-UNREHEARSED, or UNDEFENDED — and identifies failure classes the existing resilience posture is silent on. Proposes…
Agent Claude Code
Projects a system's resilience trajectory. Consumes the chaos family's present-state artifacts — the explorer's rehearsal-depth map, the analyst's load-bearing-luck register and masking map, the validator's experiment results — or discovers read-only when none are provided, and projects luck-exhaustion…
Agent Claude Code
Validates system resilience through live fault injection and failure simulation. Tests dependency failures, network disruptions, resource exhaustion, timeout handling, and recovery behavior. Executes actual chaos experiments against running systems and measures real resilience characteristics under controlled failure…
Agent Claude Code
Deep inspection for runtime correctness issues that pass compilation, linting, and tests but could fail in production. Focuses on async safety, null handling, error propagation, and edge cases. Use as FINAL gate in ship workflow. Catches the bugs that will wake someone up at 3 AM.
Agent Claude Code
Reviews code after validation passes. Proposes safe refactors for performance, structure, and maintainability without changing behavior. Must NOT introduce breaking changes unless explicitly requested. Use AFTER code-validator and test-architect pass.
Agent Claude Code
Validates code quality after implementation phases. Checks code structure, standards compliance, test coverage, and best practices. Blocks progression if critical issues found. Run after each implementation phase.
Agent Claude Code
Performs Confucian relational coherence trajectory projection on any artifact. Projects naming drift compounding, ritual decay trajectories, and relational entropy over time. Decision - HIGHCONFIDENCE/MODERATECONFIDENCE/LOWCONFIDENCE.
Agent Claude Code
Identifies what kinds of bugs the ecosystem is not catching. Analyzes the agent-taxonomy matrix to find blind spots (no agent covers a failure domain), single points of failure (one agent covers a mode), and wasteful overlap. Maps gaps to lifecycle phases and recommends new agents or workflow changes. Produces a…
Agent Claude Code
Assesses privacy compliance posture of data processing practices against applicable data protection regulations. Performs legal basis mapping, consent mechanism evaluation, cross-border transfer risk analysis, data subject rights coverage assessment, and purpose limitation checking. Multi-jurisdiction by design …
Agent Claude Code
Deep codebase exploration using multi-strategy search and relationship tracing. Combines structural pattern matching (imports, exports, class hierarchies), call chain tracing (grep-based caller/callee discovery), and targeted file reading to produce comprehensive exploration reports. No external tools required.
Agent Claude Code
Performs foundational architecture analysis on any artifact. Traces every derivation chain to its terminal foundation, classifies each termination (DEDUCIBLE, CORROBORATED, CONVENTIONAL, SMUGGLED, OPAQUE), and maps the gap between declared and actual foundational architecture. Produces a foundational architecture map…
Agent Claude Code
Validates documentation completeness and quality across all documentation surfaces. Covers API documentation (OpenAPI/Swagger), JSDoc/TSDoc coverage on public exports, changelog quality, and markdown validity. Complements public-interface-validator which focuses on README accuracy. Use for projects with significant…
Agent Claude Code
Developer advocate gate that tests real user experience. Executes README examples, validates error quality, tests exported functionality, simulates first-run scenarios. Run AFTER code-validator and test-architect pass. The "would I recommend this?" gate.
Agent Claude Code
Analyzes historical validation data across the entire ecosystem to identify patterns that correlate with score improvements. Generates data-driven recommendations for agent definition updates. This is the recursive appreciation closed-loop agent—it transforms accumulated validation history into actionable optimization…
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: