Stats
Actions
Tags
From write-for-agents
Use when asked to audit external-facing content (product page, README, docs, blog post, launch announcement, profile) for AI-agent legibility, run a compression test, check whether a differentiator survives AI summarization, or score content against the content-readiness rubric. Also fires pre-publish when content will be read by AI assistants answering trust/fit queries about the author or product.
How this skill is triggered — by the user, by Claude, or both
Slash command
/write-for-agents:content-auditThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Score a piece of external-facing content for its second reader: the AI that compresses it before any human sees it. The discipline, rubric, and templates live at the plugin root (the repo root, two levels up from this skill directory) - read them, do not paraphrase from memory:
Score a piece of external-facing content for its second reader: the AI that compresses it before any human sees it. The discipline, rubric, and templates live at the plugin root (the repo root, two levels up from this skill directory) - read them, do not paraphrase from memory:
content-readiness-rubric.md - the scored 5-dimension rubric (provable claims, claim-evidence pairing, compression survival, honest wedge, machine-clean structure)how-to-apply.md - where the rubric fits and how to run the compression testcompression-test-template.md - the compression-test worksheetworked-example.md + examples/ - calibration anchors (a 4/10 -> 10/10 rewrite; real articles scored 10/7/6)content-readiness-rubric.md fully, then the target content fully (no skimming - scoring from a skim produces inflated scores).compression-test-template.md, using its canonical question ("should I trust and use this?"): summarize the content in 3 sentences as an AI assistant would answer a buyer. Then check: does anything in those 3 sentences distinguish this from the category average? Name what survived and what got averaged away. The template is written for self-audit with external models; when auditing third-party content or when no second model is available, perform the compression yourself - and when the author's intended differentiator is unknown, treat it as "none stated" and let the test reveal whether one emerges.worked-example.md and examples/ - if your scores run higher than those anchors for similar quality, re-score.Version: 1.0.0
npx claudepluginhub kpachhai/write-for-agents --plugin write-for-agentsProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.