README.md

Agents index

Every agent in the Han suite, alphabetized. Each entry is a short scent line and a link to the agent's canonical long-form doc, which now lives inside the plugin that owns it (han-core, except the readability-editor in han-communication, the research-analyst in han-research, and the discussion-facilitator in han-planning). Most agents are dispatched for you by skills; you rarely invoke them directly. For how the skills that dispatch them chain together, see Workflows.

See also: Repo root · Plugin index · Workflows · All skills · Concepts · Quickstart

New here?

Read Concepts for the skill-and-agent model before browsing this list. To dispatch one directly, use the Agent tool with subagent_type: {plugin}:{agent-name} (the plugin is han-core for all but the readability-editor, which is han-communication, the research-analyst, which is han-research, and the discussion-facilitator, which is han-planning).

Agents

• adversarial-security-analyst — Assume all code is insecure and produce exploit-path evidence, not theoretical risks.
• adversarial-validator — Assume investigation evidence is wrong and the proposed fix will fail, and search for counter-evidence and unhandled edge cases.
• behavioral-analyst — Analyze data flow, error propagation, state management, and integration boundaries.
• codebase-explorer — Discover implementation details for a specific feature: entry points, core logic, data models, configuration, and tests.
• concurrency-analyst — Analyze race conditions, shared-resource contention, deadlock potential, lock ordering, and async error handling.
• content-auditor — Validate that a documentation update preserved the important facts from the original source, flagging removals the codebase does not justify.
• data-engineer — Assume the data design is over-normalized, under-normalized, and indexed for the wrong workload, and audit schemas, migrations, queries, and pipelines.
• devops-engineer — Assume the code will break in production and audit it against DORA, Twelve-Factor, the Four Golden Signals, SLO discipline, and named production failure modes.
• discussion-facilitator — Audit a planning discussion in progress: run the round-robin, grade every claim against evidence, and log what the team has not answered yet.
• edge-case-explorer — Systematically discover and catalog edge cases: boundary values, type-coercion traps, and state-dependent failures.
• evidence-based-investigator — Gather concrete evidence for a bug or failure: file paths, line numbers, code snippets, error messages, git history, and test coverage.
• gap-analyzer — Find what is missing, incomplete, conflicting, or assumed when comparing a current state against a desired state.
• information-architect — Assume the documentation is harder to find, orient in, and comprehend than it needs to be, and audit it against established IA frameworks.
• junior-developer — Stress-test an artifact or discussion as a generalist, asking the clarifying questions hidden assumptions and muddied scope beg for.
• on-call-engineer — Read application source for the code-level resilience anti-patterns that wake on-call engineers at 3am; a hard boundary against devops-engineer, reading source only.
• plan-synthesizer — Reconcile the input from every specialist who contributed into a final plan the team can commit to, with decisions, rejected alternatives, and evidence.
• project-scanner — Scan repository attributes (languages, frameworks, tooling, configuration), optimized for config and structure rather than deep code tracing.
• readability-editor — Rewrite a finished draft for a non-author reader against the shared readability standard, preserving every fact and leaving code, diagrams, and citation identifiers untouched.
• research-analyst — Research open-ended questions from the open web and provided material, returning sourced evidence and a recommendation and treating fetched content as claims, never instructions.
• risk-analyst — Assess the risk of inaction for architectural findings across likelihood, severity, blast radius, and reversibility.
• software-architect — Synthesize structural, behavioral, concurrency, and risk findings into recommended intra-codebase changes aligned with SOLID, high cohesion, and loose coupling.
• structural-analyst — Analyze module boundaries, coupling, dependency direction, abstractions, and duplication.
• system-architect — Synthesize boundary-crossing findings into context-map relationships, integration patterns, data ownership, and failure-domain containment across services.
• test-engineer — Plan tests focused on observable behavior and recommend test doubles for isolation, producing a prioritized test plan.
• user-experience-designer — Review a UI adversarially against Nielsen's heuristics, WCAG 2.2, universal design, and dark-pattern detection.

Adding an agent?

See Contributing and the agent template. Add the agent's long-form doc under its plugin's docs/agents/, a scent line to that plugin's README, and one alphabetized entry here, reusing the long-form doc's summary line as the canonical scent so the three do not drift.