AI & Agents Agent Skills

Explicit activation required: do not invoke or load this skill from an implicit match; wait for explicit user or operator invocation or a source-authorized parent-route selection. Prefer simulation, preview, or inspect-only paths before any real mutation, then require one explicit confirmation for the concrete mutating step. Use when the action can be previewed and mistakes would be costly, such as delete, restore, migrate, or reconfigure operations. Do not use for purely analytical tasks or when the primary question is whether execution is authorized at all.

Explicit activation required: do not invoke or load this skill from an implicit match; wait for explicit user or operator invocation or a source-authorized parent-route selection. Design a bounded local eval suite, prompt set, report, or validator plan after selection finds no adequate existing surface, using deterministic checks first and trace or rubric review only when needed. Use when a local eval needs positive, negative, collision, or regression cases around a clear behavior or invariant. When the parent has classified design but the explicit invariant or behavior, target repository, local owner path, or rejected existing surfaces is unavailable inside the active evidence boundary, stop with blocked_missing_input; do not relabel missing input as deferred_owner_boundary. Do not use when an existing eval should simply run, when only a local intake packet is needed, or when central aoa-evals proof doctrine is the target.

Explicit activation required: do not invoke or load this skill from an implicit match; wait for explicit user or operator invocation or a source-authorized parent-route selection. Inspect repo-local eval ports, tests, validators, scripts, reports, intake files, and central aoa-evals standards before new eval work, then select the smallest existing surface or a no-fit route. Use when the task asks which eval should apply to a change, failure, regression, or trigger miss. When the parent has classified selection but the target repository, touched paths, local eval port, nearby proof surfaces, or central standard is unavailable inside the active evidence boundary, stop with blocked_missing_input; do not relabel missing input as deferred_owner_boundary or infer fit/no-fit. Do not use when the eval is already selected and should run, when no-fit has already been established, or when the real issue is source authority rather than eval coverage.

Create OpenCode plugins using the opencode-ai/plugin SDK. Use when user wants to build a plugin, extend OpenCode, intercept tool execution, add custom tools, react to events, create a hook, block commands, add custom auth, or add automation to OpenCode. Also use for 'opencode plugin', 'custom tool', 'tool hook', 'plugin hook', 'logging for all tool calls', 'restrict what commands the AI can run', 'lifecycle hooks', 'middleware', 'tool.execute.before', 'tool.execute.after', 'auth hook', 'chat.params', 'intercept bash commands', 'block file writes', 'add audit logging', 'custom model provider auth', 'inject context before LLM call', 'fire-and-forget background task', 'watch session events'.

Create and refine OpenCode agents via guided Q&A. Use when creating an agent file, improving an existing agent, configuring AGENTS.md frontmatter, setting subagent_type, defining allowed-tools, restricting permissions, designing agent routing, writing a system prompt, configuring mode: subagent, setting a permission block, or working with the Task tool delegation pattern. Examples: 'Create an agent for code reviews', 'My agent ignores context', 'Add a database expert subagent', 'Make my agent faster', 'Agent triggers on wrong requests'.

Evaluate, score, and remediate agent skill collections using a 9-dimension quality framework (Knowledge Delta, Mindset, Anti-Patterns, Specification Compliance, Progressive Disclosure, Freedom Calibration, Pattern Recognition, Practical Usability, Eval Validation). Performs duplication detection, generates remediation plans with T-shirt sizing, enforces CI quality gates, validates artifact conventions, tracks score trends, and ensures tessl registry compliance. Use when evaluating skill quality, auditing SKILL.md files, scoring agent skills, generating remediation plans, detecting duplicate skills, validating skill format, enforcing quality gates, optimizing for A-grade publication, comparing audit baselines, batch skill assessments, or checking tessl compliance. Triggers: 'check my skills', 'skill audit', 'improve my SKILL.md', 'quality check', 'A-grade scoring', 'quality gates', 'eval validation', 'audit all skills', 'remediation plan', 'skill judge', 'dimension scoring'.

Ready up Jira tickets for refinement by gathering context from linked incidents, populating a YAML template following the project's 'ready for refinement' standard, validating against a JSON schema, and generating a structured markdown document. Use when asked to 'ready up', 'ready a ticket', 'prepare for refinement', or 'write up PROJ-NNN'. Keywords: jira, ticket, refinement, ready, backlog, YAML template, JSON schema, context, conditions of satisfaction, acceptance criteria.

Execute a wave-based plan by spawning parallel agents per wave, enforcing merge gates, and building self-contained per-agent prompts from wave document rows. Use when asked to: run the plan, execute the plan, start the waves, launch agents, execute wave N, run wave execution. Reads a wave document produced by wave-execution-planner.

Use when the user reports a failure, flaky test, CI error, runtime log, crash, UI symptom, regression, or suspected bug where a repro, hypotheses, instrumentation, browser/CI evidence, or human observation must decide root cause before a fix is safe.

Use when an accepted plan, checklist, approved scope, or known root-cause fix should be implemented, continued, resumed, or verified with focused action.

Review a saved standards document via a one-standard-at-a-time interview. Prompts the user for the file path, parses individual standards, then presents each one for evaluation with concrete options (Accept / Revise / Reject / Defer). Produces a summary report at the end.

Route AoA eval-lane work by raising the session readiness packet, checking local and central eval surfaces plus any repo-family or workspace port inventory, then selecting exactly one apply, local-need, design, or session-mining child while preserving owner boundaries. Use when work concerns existing or new evals, local ports, aoa-evals, validators, tests, scripts, traces, regressions, trigger misses, MCP candidate packets, or session readiness as evaluation surfaces. An explicit request to design a bounded local eval suite selects aoa-eval-design even when invariant or target inputs are missing; that child blocks for them instead of rerouting to select. Before acting, read this root skill to EOF and fully read exactly one selected child; a bounded prefix is not a complete load. Do not use for ordinary one-off tests, source-of-truth mapping, decision records, memo writeback, or direct central proof-authority edits.

Audit your Claude Code setup against prompt caching best practices. Checks ordering, tool stability, dynamic content handling, and hook injection patterns. Returns a scored report with specific fixes.

Route AoA decision-lane work by classifying the task, then select and fully read exactly one find, create, or correct child before that child performs graph lookup or a write, while keeping repo-local decision files authoritative. Use when the task mentions docs/decisions, decision indexes, decision graph, ADRs, changed paths, source-surface rationale, supersession, or cross-repo decision symmetry. Do not stop at the root after naming a child. Do not use for ordinary docs edits, unresolved source-of-truth mapping, or trivial changes that do not need durable rationale.

Use the aet CLI as an evidence-driven Agent engineering quality and control layer. Audit instructions, review intent against diffs, trace explicit proofs, diagnose structured failures, stage regression candidates, or run bounded evidence-gated asset evolution. Works with any Agent that can read instructions and run a local CLI.

Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and excessive conjunctive phrases. Credits: Original skill by @blader - https://github.com/blader/humanizer

Create and edit Obsidian Bases (.base files) with views, filters, formulas, and summaries. Use when working with .base files, creating database-like views of notes, or when the user mentions Bases, table views, card views, filters, or formulas in Obsidian.

Search Google Scholar for academic papers and author profiles. Use when discovering papers to triage — by keyword, author, or year range. Returns titles, authors, abstracts, and links. Triggers: search papers, find papers, google scholar, search scholar, find author, author profile, literature search, paper discovery, search academic literature.

Triage an academic paper into a structured reference summary for the research repo. Use when given an arxiv ID, arxiv URL, DOI, or paper PDF. Creates references/{slug}.md; adds REVIEWED.md entry and REFERENCE_INDEX.md row. Triggers: triage paper, add paper, analyse paper, review paper, literature review, survey paper, benchmark paper, deep dive paper, arxiv, DOI.

Triage a tool or library into a structured reference summary for the research repo. Use when given a GitHub URL, npm package, PyPI package, or CLI tool name. Creates references/{slug}.md with architecture overview and scope assessment; adds REVIEWED.md entry and REFERENCE_INDEX.md row. Triggers: triage tool, add tool, analyse tool, GitHub project, npm package, PyPI package, library, framework, CLI tool, MCP server, open source project.

Token-efficient MCP resource ladder for metagit workspaces — catalog, map, layered prompts, project/repo drill-down. Use when the IDE host exposes metagit resources/read instead of shell CLI.

Ongoing workspace and project management for OpenClaw and Hermes agents. Use when starting work, organizing repos, or before creating a new project folder so existing metagit projects are reused instead of duplicated.

Orchestrate a reference-implementation rewrite across source and target repos using campaigns, parity registry conventions, objectives, and subagent handoffs.

Configure shared coordination state (objectives, handoffs, approvals, events) across multiple agents and machines via METAGIT_STATE_URL and the ops HTTP backend. Use when Hermes subagents, CI runners, or humans must see the same objective queue without Syncthing JSON files.