review-algorithm-plan Agent Skill

Review recommendation, feed, ranking, reranking, mixing, recall, and search algorithm plans against source-grounded industrial evidence. The Skill separates user-provided facts, public production evidence, inference, and unverified assumptions, then turns the review into falsifiable experiments, launch gates, guardrails, and rollback conditions.

Install

npx skills add https://github.com/carrieluu/review-algorithm-plan-agent-skill --skill review-algorithm-plan

For a global Codex installation without prompts:

npx skills add https://github.com/carrieluu/review-algorithm-plan-agent-skill --skill review-algorithm-plan --agent codex --global --copy --yes

The skills CLI documents that it collects anonymous telemetry by default, including the skill name, skill files, and a timestamp. To opt out for the install command:

DISABLE_TELEMETRY=1 npx skills add https://github.com/carrieluu/review-algorithm-plan-agent-skill --skill review-algorithm-plan

Review-first installation prompt

Install this Agent Skill for me.

Skill page: https://github.com/carrieluu/review-algorithm-plan-agent-skill
Source URL: https://github.com/carrieluu/review-algorithm-plan-agent-skill/tree/main/skills/review-algorithm-plan
Skill name: review-algorithm-plan
Creator: carrieluu
Preferred install command: npx skills add https://github.com/carrieluu/review-algorithm-plan-agent-skill --skill review-algorithm-plan

Please open the skill page and source, review SKILL.md plus any companion files, and explain anything risky before installing.

If shell commands are available, prefer the install command above. If you install manually, copy the complete skill directory that contains SKILL.md, including scripts, references, assets, agents, and any other files shown on the skill page. Preserve the relative folder structure. Do not install SKILL.md alone. After installing, verify the target skills folder contains SKILL.md and all companion files needed by this skill.

Package contents

skills/
└── review-algorithm-plan/
    ├── SKILL.md
    ├── agents/
    │   └── openai.yaml
    └── references/
        └── industry-rag-playbook.md

The Skill is instruction-only. It contains no scripts, executables, binary assets, symlinks, credentials, fixed network endpoints, or bundled software dependencies.

Security and privacy notes

• The Skill asks the agent to search public sources when an algorithm-plan judgment depends on industrial evidence. It does not grant network access itself; the receiving agent must already provide suitable search or browsing tools.
• The Skill can inspect user-provided internal plans before public comparison. Do not copy confidential plan text, internal identifiers, metrics, user data, or credentials into external search queries.
• Search queries and opened pages may be processed by the receiver's search, browser, or connector provider. Review that provider's privacy and retention terms before using sensitive material.
• Public production reports are often operator-reported and may not be independently audited. The Skill requires explicit evidence grades and transferability checks, but users should still inspect the cited sources before making launch decisions.
• If retrieval is unavailable, the Skill requires the agent to label the review 未完成外部验证; it must not present remembered examples as confirmed-current production evidence.
• The install command follows the current repository state unless the receiver pins a tag or commit. Security-sensitive users should install or review an immutable release.

License

MIT. See LICENSE.

Marketplace listing

The GitHub repository is the canonical Skill page until a marketplace listing exists. A public GitHub repository does not guarantee automatic listing by any particular marketplace. When a SkillsMP or skills.sh page is available, add its URL to the review-first installation prompt without changing the source URL.