agents-scripts

Plugin agents-scripts from marketplace agents-scripts in ayaqen/Agents-Scripts. Commands read from the repo's manifest — 1 month ago.

Skills (17)

• ci-green

  Drive a failing CI pipeline back to green: diagnose, fix, re-kick, repeat until merged.

• context-budget

  Work in large codebases without drowning context: targeted reads, delegated searches, externalized state.

• debug-loop

  Hypothesis-driven debugging: reproduce, isolate, prove the cause before writing the fix.

• dependency-vet

  Evaluate a new dependency before adding it: health, security, cost, and the no-dependency alternative.

• eval-design

  Design evals for an LLM feature: graded cases, honest metrics, and a regression set that gates changes.

• model-tiering

  Split work across model tiers: a frontier model orchestrates and reviews, cheaper subagent models execute well-specified tasks.

• pr-review

  Review someone else's pull request: correctness, scope, tests, and an actionable verdict.

• prompt-triage

  Debug LLM prompt or agent misbehavior systematically: isolate inputs, form hypotheses, test one variable at a time.

• regression-fix

  Fix a bug by writing the failing regression test first, then making it pass.

• release-checklist

  Cut and verify a software release: changelog, tag, publish, and proof it actually shipped.

• repo-onboarding

  Map an unfamiliar codebase before making changes: layout, entry points, conventions, hot spots.

• safe-refactor

  Behavior-preserving refactoring in small verified steps, with a rollback path at every point.

• security-review

  Security pass over a diff, feature, or dependency: trace untrusted input, verify exploitability, fix the sink.

• self-review

  Review your own diff like a hostile reviewer before committing or opening a PR.

• session-handoff

  Write a handoff note so the next agent session or teammate can continue without re-discovery.

• skill-author

  Create or update a skill in this repository: scaffold, write, validate, sync.

• tool-design

  Design tools and MCP servers agents call correctly on the first try: schemas, descriptions, outputs, errors.

Subagents (5)

• bug-hunter

  Reproduces and root-causes a failing test or reported bug via hypothesis-driven debugging; returns the proven cause, not a guess.

• code-reviewer

  Reviews a diff or PR for correctness, scope, tests, and consistency; read-only; returns blocking and non-blocking findings.

• mechanical-editor

  Executes precisely-specified mechanical edits — exact string replacements, file writes from verbatim content, bulk renames. No judgment calls.

• security-auditor

  Adversarial security pass over a change or codebase; traces untrusted input to sinks and confirms exploitability before reporting.

• test-writer

  Writes failing-first regression tests for a described bug or uncovered behavior, matching the repo's existing test conventions.

Similar plugins