Agent Skills

Personal skills and workflows for Codex and GPT agents.

Installation

npx skills add plimeor/agent-skills

Install a single skill:

npx skills add plimeor/agent-skills --skill ops-url-reader

Project Structure

• skills/<skill-name>/SKILL.md: each skill has its own directory, and SKILL.md is the entrypoint.
• The name: field in SKILL.md frontmatter must match the parent directory name exactly.
• README.md is the public index. Update it whenever a skill is added, removed, or renamed.

Skills

Skills are grouped by primary mode.

Code

• code-scope-gate: Scope gate before coding; prevent over-implementation and converge on the smallest correct change.
• code-standards-gate: Review specs, diffs, and implementation boundaries against personal code standards, with emphasis on public contracts, persisted state, and unnecessary abstractions.
• code-test-strategy: Test-strategy gate for coding tasks; avoid test-driven production complexity, premature tests, and implementation-detail tests.

Decision

• decision-look-before-leap: Look-before-you-leap check before answers, recommendations, or decisions; use independent subagent scrutiny when context-inertia risk is high.

Knowledge

• knowledge-project-docs-maintenance: Maintain project docs layering, language, naming, and living-spec cleanup strategy.

Meta

• meta-code-standards-calibration: Calibrate code-standards-gate against human PR/MR reviews and decide whether rules belong in a global skill, project rules, tooling, or local context.
• meta-context-engineering-global: Context engineering for global rules files, focused on rules that remain valid across tasks, projects, and sessions.
• meta-context-engineering-project: Project-level context engineering for repo-local rules, task context packing, and project-specific drift diagnosis.
• meta-gpt-prompt-maintenance: Maintain and upgrade GPT-oriented prompt artifacts, including SKILL.md, AGENTS, system/developer prompts, agent workflows, evals, and grader prompts.
• meta-subagent-orchestration: Orchestrate focused sub-agents for parallel investigation, delegated implementation, and verification while keeping the main agent responsible for integration.

Ops

• ops-bear: Read, search, create, edit, organize, and open Bear notes through the local Bear App CLI.
• ops-codex-session-maintenance: Maintain local Codex session state: inspect and back up first, then archive old sessions/worktrees, rotate logs, and generate handoffs.
• ops-url-reader: Extract main content from arbitrary URLs through defuddle.md.

Writing

• writing-blog: Create and improve blog posts with the SCQA method.
• writing-blog-illustration: Generate illustration prompts for blog posts, especially workflow, architecture, and abstract-concept visuals.
• writing-humanizer: Reduce AI-writing traces so generated docs and drafts read more naturally and human-authored.
• writing-reader-feedback: Simulate a specified reader reading an article section by section and report raw reading-experience feedback.