corporate-espionage
Skill0 starsChanged 3 months agoMITView on GitHub
Reverse-engineer a single feature from source code into a formal spec sheet. The output is language-neutral and contains no reproduced code, making it suitable for cleanroom reimplementation, architectural review, or onboarding. The user must identify the feature source. Accepted forms: If no source is provided, ask for it before proceeding. One focused question only. Run these phases in strict order. Do not begin Phase 2 before Phase 1 is complete. 1. Resolve the source: - If path: read the file…
In this repository24 other files it gives its agents.
- dispatching-parallel-agentsSkill
- epub-readerSkill
- executing-plansSkill
- finishing-a-development-branchSkill
- github-raw-readerSkill
- magisk-moduleSkill
- notepadpp-plugin-skillSkill
- planningSkill
- receiving-code-reviewSkill
- requesting-code-reviewSkill
- reviewSkill
- skill-creator2Skill
- subagent-driven-developmentSkill
- systematic-debuggingSkill
- technical-writerSkill
- test-driven-developmentSkill
- text-humanizerSkill
- using-git-worktreesSkill
- using-superpowersSkill
- verification-before-completionSkill
- windows-diagnosticsSkill
- writing-plansSkill
- zed-extensionSkill
- zed-themeSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
20 rules. About 0 are common, 17 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈17
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 31
If no source is provided, ask for it before proceeding.
- Distinctive: line 35
Run these phases in strict order.
- Distinctive: line 40
If path: read the file
- Distinctive: line 41
If GitHub URL: fetch raw content
- Distinctive: line 42
If pasted code: use directly from context
- Distinctive: line 53
If the boundary is ambiguous (multiple candidates exist), list them and ask the user to confirm.
- Very specific: line 58
Read references/analysis-agent.md before executing this phase.
- Distinctive: line 62
Input catalog: - every value, resource, or signal entering the feature
- Distinctive: line 63
Output catalog: - every value, artifact, or side effect the feature produces
- Distinctive: line 64
Process map: - every logical transformation from inputs to outputs, described in plain language with no reproduced code
- Very specific: line 74
Use the extraction results from Phase 2 to produce the spec sheet.
- Very specific: line 75
Use the template in references/output-template.md exactly.
- Distinctive: line 82
Before presenting the output, verify:
- Distinctive: line 84
Every input in Phase 2 maps to at least one FR or NFR
- Distinctive: line 85
Every process in Phase 2 maps to at least one FR
- Distinctive: line 86
No code is reproduced anywhere in the spec (variable names in I/O tables are acceptable.
- Distinctive: line 91
If any check fails, fix the gap before presenting.
- Distinctive: line 95
Read only what is within the stated feature boundary
- Distinctive: line 96
Do not pull in the entire codebase unless the feature has no meaningful boundary without broader context (state this explicitly if so)
- Distinctive: line 99
If the source is inaccessible, report the failure and wait for the user to provide an alternative
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Feature Spec
Reverse-engineer a single feature from source code into a formal spec sheet. The output is language-neutral and contains no reproduced code, making it suitable for cleanroom reimplementation, architectural review, or onboarding.
Entry Condition
The user must identify the feature source. Accepted forms:
- A file path (local or uploaded at
/mnt/user-data/uploads/) - A GitHub URL (raw or standard; use
github-raw-readerskill if loaded) - A pasted code block in the conversation
- A function, class, or module name inside a larger file already in context
If no source is provided, ask for it before proceeding. One focused question only.
Workflow
Run these phases in strict order. Do not begin Phase 2 before Phase 1 is complete.
Phase 1: Source Resolution and Feature Boundary
- Resolve the source:
- If path: read the file
- If GitHub URL: fetch raw content
- If pasted code: use directly from context
- Identify the feature boundary. The boundary is the smallest cohesive unit
that delivers the described behavior. It may be:
- A single function or method
- A class with its public interface
- A module or file if the user names it explicitly
State the boundary explicitly before continuing:
- Feature name
- Location (file, lines if applicable)
- One-sentence purpose
If the boundary is ambiguous (multiple candidates exist), list them and ask the user to confirm. Do not guess.
Phase 2: Analysis Extraction
Read references/analysis-agent.md before executing this phase.
Extract three catalogs from the resolved source:
- Input catalog - every value, resource, or signal entering the feature
- Output catalog - every value, artifact, or side effect the feature produces
- Process map - every logical transformation from inputs to outputs, described in plain language with no reproduced code
Record all assumptions made during extraction. Surface uncertainties explicitly rather than guessing.
Phase 3: Spec Sheet Generation
Read references/spec-writer.md before executing this phase.
Use the extraction results from Phase 2 to produce the spec sheet.
Use the template in references/output-template.md exactly.
The spec sheet must stand alone. A second agent with no access to the original source should be able to reimplement the feature using only this spec.
Phase 4: Self-Check
Before presenting the output, verify:
- Every input in Phase 2 maps to at least one FR or NFR
- Every process in Phase 2 maps to at least one FR
- No code is reproduced anywhere in the spec (variable names in I/O tables are acceptable; code blocks are not)
- Assumptions and open questions are recorded
- The Cleanroom Firewall Notice is present at the end
If any check fails, fix the gap before presenting.
Source Reading Rules
- Read only what is within the stated feature boundary
- Do not pull in the entire codebase unless the feature has no meaningful boundary without broader context (state this explicitly if so)
- For large files, identify the relevant section first, then read only that section
- If the source is inaccessible, report the failure and wait for the user to provide an alternative
Output
Present the completed spec sheet inline as a Markdown document.
Also save it to /mnt/user-data/outputs/feature-spec-<feature-name>.md and
call present_files.
Reference Index
references/analysis-agent.md- extraction rules for inputs, outputs, processesreferences/spec-writer.md- requirement writing rules and quality gatesreferences/output-template.md- exact spec sheet templatereferences/design-log.md- design decisions for this skill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

