b-mendoza/agent-skills
23 skills2.5k installs23 starsGitHub
Install
npx skills add https://github.com/b-mendoza/agent-skillsSkills in this repo
1Validate Implementation PlanThe validate-implementation-plan skill orchestrates a safe standalone audit of AI-generated or human implementation plans. Untrusted PLAN_PATH is read only by plan-snapshotter; later stages work from SNAPSHOT_PATH, numbered requirements, approved local evidence, structured findings, and summarized user answers. Required inputs are PLAN_PATH and ORIGIN_CONTEXT (or a baseline question first); optional OUTPUT_PATH defaults to sibling .audit.md and SNAPSHOT_PATH to .audit-input.md. SOURCE_CONTEXT_PATHS allow-list local files classified as baseline-context, local-technical-evidence, mixed, or unreadable. Pipeline phases cover intake, snapshot sanitization, requirements extraction, evidence review, and findings with severities critical, warning, and info. Output contract returns compact AUDIT: PASS|FAIL|BLOCKED|ERROR handoff with finding counts unless the user requests the full report. Use when reviewing implementation plans, design proposals, task breakdowns, or architecture outlines for audit-only feedback.1.6kinstalls2Recency Guardrecency-guard from b-mendoza/agent-skills implements guardrails so coding agents verify information freshness before responding. It reduces outdated API guidance and deprecated pattern recommendations during pre-ship review of agent behavior.115installs3Pr CreatorThis skill creates review-ready pull or merge requests from the current branch. It routes six focused specialists through repository topology, preflight, diff analysis, drafting, reviewer/label metadata, and submission, and it creates nothing until the user approves the exact preview. A developer uses it when a branch is ready for review to open a PR safely. It is fork-aware and idempotent, so an existing open PR for the same head and base stops creation.106installs4Generate Handoff DocumentGenerate Handoff Document produces a resumable handoff package from an in-progress conversation, review, debugging session, or investigation. A developer uses it to save work so a fresh agent can continue without the chat history. It writes one human-readable handoff document plus sibling structured artifacts, with transcript materialization, update-mode backups, and bounded review repair.88installs5Clarifying AssumptionsThis skill runs the conversational clarification layer for a workflow-orchestration pipeline. In upfront mode it challenges a whole plan before execution starts; in critique mode it challenges one task just before it runs. A developer uses it to surface open assumptions and record clarification decisions, delegating artifact analysis and file writes to bundled subagents.74installs6Committing Scoped ChangesThis skill orchestrates scoped git commits from an explicit set of file or folder paths after the user asks to commit. It protects the requested path boundary, plans atomic commit groups, verifies staged diffs, and executes one approved commit at a time. A developer uses it to split broad changes into a clean review series while preserving unrelated work.65installs7Review Pull RequestThis skill reviews a single pull request end to end. It collects PR context, surfaces evidence-backed defects and residual risks, drafts GitHub-ready comments, verifies the review package, and writes a local review file, optionally posting the exact approved review. A developer uses it to audit a PR or prepare request-changes feedback. It prefers fewer, stronger findings and treats every finding as provisional until a verifier passes.65installs8Responding To Pr Review CommentsThis skill assesses and responds to pull request review comments. It collects the comment inventory, evaluates each piece of feedback with evidence, drafts replies, and writes a verified local report, then optionally posts the exact approved replies to supported GitHub review-comment threads. A developer uses it when a PR has review feedback that needs evidence-based responses. It treats review comments as proposals to evaluate, not commands to accept, and never edits code.58installs9Refactoring CodeThis skill coordinates behavior-preserving refactors of existing code. It runs a pipeline that maps current behavior, plans the smallest change, implements only the approved plan, and reviews the diff, asking before mutation unless auto-approved. A developer uses it to simplify, split, rename, or move code without changing what it does. It refuses to cross protected surfaces and treats validation as a contract with recorded evidence.37installs10Rewriting Code StrictlyThis skill rewrites existing Python, TypeScript/JavaScript, or Go code to be strictly typed, boundary-validated, and more maintainable while preserving behavior. It coordinates a baseline map, a strategy, an approved implementation, and a review, fetching a language playbook just in time. A developer uses it to harden code, remove unsafe escape hatches, add validation, or align with mypy, tsc, or go vet. It records static-typing versus runtime-validation decisions and the validation commands it ran.35installs11Diagnosing Root CausesThis skill diagnoses runtime bugs, crashes, regressions, failing CI/CD pipelines, and underspecified user reports through read-only, evidence-first root-cause analysis. It classifies the issue, gathers a cited evidence base, produces a traceable RCA draft, and reviews it for grounding and safety before delivering. A developer uses it to find the root cause of a reported problem without the skill executing risky fixes.34installs12Improving Test SuitesImproving Test Suites is a portable orchestrator that turns a named test suite into the smallest useful behavior-focused harness. A developer uses it to trim, rewrite, delete, or harden tests around public contracts, business logic, schemas, security behavior, and edge cases. It gates every destructive change before mutation, verifies approved behavior coverage survived, and returns exactly one named handoff status.33installs13Generate Flow DiagramGenerate Flow Diagram turns workflow descriptions into reviewed Markdown documents that each contain one Mermaid flowchart. A developer uses it to create, refine, repair, or decompose process flows, agent operating procedures, and human-in-the-loop gate maps. It runs an orchestrator that classifies the run mode, keeps approvals, stages writes, and only returns or writes candidates after independent review.28installs14Prompt StructurerThis skill turns prose prompts into executable XML prompt contracts through staged analysis passes. It decomposes source intent, separates philosophy from constraints, surfaces implicit behavior and anti-patterns, and builds observable success criteria before assembling the final XML prompt. A developer uses it when a natural-language prompt needs to become a reliable, drift-resistant agent contract. It selects the smallest deterministic flow and returns auditable assembly notes.27installs15Workflow Skill ArchitectWorkflow Skill Architect turns repeatable workflows, prompts, or existing skill packages into portable agent-skill artifacts or review reports. A developer runs it when creating, extending, refactoring, or reviewing skills for Claude Code, Cursor, or OpenCode. It classifies the request, stages generated files in a scratch directory, and only writes to a real package after explicit approval. It also emits a canonical review report against a defined schema.19installs16Council Of AdvisorsThis skill runs a structured nine-seat council deliberation on an idea, project, business, startup, or goal. It dispatches independent mental-model seats, classifies the decision as reversible or irreversible, synthesizes a recommendation with dissent and a kill criterion, and writes a decision handoff file. A developer or founder uses it to stress-test a decision and get multiple adversarial and optimistic perspectives.14installs17Fetching Work ItemFetching Work Item retrieves a Jira ticket or a GitHub issue into a docs/<KEY>.md file as a read-only, validated Markdown snapshot. A developer uses it as the Phase 1 fetch step of a larger work-item orchestration workflow, either from an orchestrator or directly. It detects the platform from the input and loads the matching playbook, then stops after producing the snapshot and a structured handoff.10installs18Orchestrating WorkflowOrchestrating Workflow coordinates an end-to-end Jira or GitHub work-item workflow from initial fetch through per-task implementation. A developer supplies a Jira ticket URL, a GitHub issue URL, or coordinates, and the skill detects the platform and loads the matching playbook. It drives a seven-phase pipeline (fetch, plan, critique, create child items, then plan and execute each task) while staying a routing layer that dispatches to downstream skills.10installs19Planning Codebase RestructuringPlanning Codebase Restructuring produces an evidence-backed restructuring plan for an existing codebase without implementing any file moves or refactors. A developer uses it to map the current architecture read-only, run Domain-Driven Design and Screaming Architecture analysis, and get a persisted decision report. It protects the target codebase from mutation and routes analysis through subagents, writing only the final report.8installs20Refine TaskThis skill reviews a single Jira or GitHub work item for implementation readiness. It dispatches one reviewer that produces exactly one tracker-facing refinement comment, either as a draft or, after approval gates pass, posted to the tracker. A developer uses it to refine a ticket, issue, or epic before it enters implementation. It is reviewer-only and never edits tracker metadata, bodies, labels, or child work.4installs21Review Software Engineer CvThis skill reviews and tailors a software engineer resume against a specific job posting. It normalizes the CV and posting, maps job requirements to CV evidence, rewrites bullets truthfully, and validates that recommendations stay interview-defensible. A developer uses it to align their experience to a role, improve hiring-manager appeal, and tune ATS-readable wording. It keeps private candidate material local and never submits it to external resume scanners.3installs22Analyzing Recent Project Stateanalyzing-recent-project-state is a skill that produces a verified, read-only snapshot of recent repository work from local Git evidence. It resolves a comparison base, collects bounded git evidence, drafts a developer-facing snapshot, verifies it, and returns either a Project State Snapshot report or an escalation envelope. A developer uses it to see what changed recently, whether a branch is ready, and what risks remain before continuing, reviewing, merging, or handing off. The run is read-only end to end and converts any mutation request into a reported risk.2installs23Improving Skill DefinitionImproving Skill Definition is a portable orchestrator that falsifies and then repairs an existing first-party agent skill package. A skill author uses it to audit workflow coherence, subagent architecture, contracts, personality, package hygiene, and prompt sufficiency. It stops for explicit in-run approval, applies only approved edits, and validates closure against a baseline diff.2installs