Skip to main content
Need help?
Ask in DiscordOpen a GitHub issueSupport guide
This directory collects Codex-compatible skills that are useful for everyday workflows and personal operations. Each skill should be usable as an installable package with a human-facing README.md, an agent-facing SKILL.md, and only the supporting resources it needs.

​
Current Skills

SkillPackage pathWhat it demonstrates
Agent Runtime Cache Benchmarkskills/agent-runtime-cache-benchmark/README.mdLocal-first comparison of cold and warm agent runs, cache-break detection from structured run artifacts, and report-driven prompt-cache tuning
Daily News Watcherskills/daily-news-watcher/README.mdSQLite-backed persistent news monitoring, RSS/Atom + optional Playwright fetching, URL+hash deduplication, and per-run Markdown reports
Garbage Collectorskills/garbage-collector/README.mdPreview-first local cleanup, readable CSV rules, reversible duplicate cleanup, and explicit approval before destructive actions
Local Document Organizerskills/local-document-organizer/README.mdPreview-first folder organization, readable classification rules with confidence scoring, reversible moves with action logs, and a no-delete, no-overwrite safety contract
Personal Knowledge Captureskills/personal-knowledge-capture/README.mdLocal-first watch-path registration, incremental source detection, SQLite state, and cited Markdown note generation
Prompt Cache Agent Harnessskills/prompt-cache-agent-harness/README.mdClaude prompt-cache layer planning, captured usage metadata review, cache-read/write reporting, and optional caller-supplied input-cost estimates
Safety Escalation Reviewskills/safety-escalation-review/README.mdLocal evidence review, redacted escalation memos, and human-owned safety handoff checklists without external reporting

​
Skill Package Shape

Each shared skill should keep the package itself small: 
skills/<skill-name>/
  README.md
  SKILL.md
  agents/openai.yaml
  scripts/
  references/
Use README.md for the human-facing story: why the package exists, who it is for, and the end-to-end workflow a student or operator should understand before using it. Use SKILL.md for the Codex invocation contract. Use scripts/ for deterministic helpers, references/ for rule files or supporting guidance, and agents/openai.yaml for UI metadata.

​
Contribution Type

Practitioner skill packages are code-plus-docs contributions. A package should include a human-facing README.md, an agent-facing SKILL.md, agent metadata, deterministic scripts, reference rules, and the minimum supporting documentation needed to make the skill reviewable and runnable. Do not treat these issues as documentation-only unless the requested change only edits explanatory pages.

​
Contribution Type

Practitioner skill packages are code-plus-docs contributions. A package can include Markdown instructions, agent metadata, deterministic scripts, reference rules, and the minimum supporting documentation needed to make the skill reviewable and runnable. Do not treat these issues as documentation-only unless the requested change only edits explanatory pages.

​
Practitioner Boundary

Practitioner skills should show where Codex execution is stronger than a plain chat answer: local files, persistent rules, repeatable workflows, browser or system interaction, and explicit safety checks.