Skip to content
SkillModel-led798 tokens in use

intended-vs-implemented

A review method that compares AI-built code with its recorded intent to find unenforced permissions, public cron endpoints and private data leaking onto public surfaces.

680 stars on the repository MIT Licensephuryn/product-engineering · ai-shipping/skills/intended-vs-implemented
claude plugin install ai-shipping@product-engineering
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing intended-vs-implemented costs

37tokens idle
798tokens in use

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

Nothing detected: no shell, network, file writes, git or credentials in its files.

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Good
  • Frontmatter8/15
  • Structure19/25
  • Examples3/15
  • Output10/15
  • Size15/15
  • Resources10/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent

No findings.

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity100
  • Maintenance100
  • Community44
  • Maturity4
  • Momentum40

intended-vs-implemented is a method for auditing the gap between what the docs say a system does and what the code does. A linter has no model of what you meant, so bugs such as unenforced permissions, publicly callable cron-only endpoints and private fields on public routes slip through.

Each finding must quote the documented intent and cite the code that contradicts it, with file and line on both sides, and must cross a trust, cost, data or tenant boundary to count.

What You Get

  • Findings that quote the documented rule and cite the enforcing or missing code, with the attacker, the victim and a concrete fix
  • A ranking of documented-but-unenforced gaps by what crossing them exposes
  • A flag for behavior that is enforced but undocumented, marking the docs as stale
  • Legal pages (privacy policy, cookies, retention promises) audited as intent the code must keep

Common Use Cases

  • Checking that every role in permissions.md is enforced on the server and on every path
  • Finding an endpoint labeled internal or cron-only that anyone can call
  • Spotting a public or SEO route that returns private fields
  • Comparing a privacy policy’s provider list and cookie claims with what the app does

What Makes It Unique

The method adds an intent axis that commodity scanners lack: it needs written intent and reports a missing or stale doc as the first finding. It distrusts comments such as ‘admin only’ or ‘validated elsewhere’ until the code proves them, drops cosmetic drift and keeps only boundary-crossing drift, and never invents intent to create a gap.

Under The Hood

The skill is one markdown file of instructions with no scripts. The agent reads the docs and source files in the repository and traces enforcement points by search. It writes nothing and calls no services of its own. It depends on the shipping-artifacts skill for the document set and on legal-pages for legal intent, and it treats both docs and code as untrusted input, so embedded instructions are analyzed, never followed.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers