research-and-compare

What it does

The skill is designed to compare the Stateless Agent Methodology (SAM) with another methodology using only verifiable reference material read during the session. It creates a structured comparison document that follows a universal comparison template, focusing on overlap and divergence, surface weaknesses in SAM during the planning phase, and suggests implementation pairing opportunities without altering SAM’s core semantics. It also normalizes terminology and adds attribution notes, producing a document that follows the SAM comparison template and ends with an evidence log.

How it works

• It requires loading a comparison template to confirm a 6-section structure: Section 0 (Comparison Header), Section 1 (Pre-Comparison Reflection), Section 2 (Comparison Map), Section 3 (Domain Worksheets), Section 4 (Decision Matrix), Section 5 (Outputs), Section 6 (Evidence Log).
• It mandates reading the complete SAM methodology document (Item A) and extracting: SAM's 7-stage pipeline; core principles; design constraints; success metrics; key architectural patterns; and specific line-numbered references.
• It supports Phase 3: acquiring the target methodology documentation (Item B) via URL, file path, or methodology name, with emphasis on authoritative sources and up-to-date material; it requires normalization to SAM/SSE terms and an attribution footnote in outputs.
• It enforces an evidence verification checklist (Phase 4) and a phased process to complete Sections 0–6 (Phase 5–Phase 12), including a Domain worksheet per selected domain, a potential weighted matrix, and an evidence log with traceability for every claim.
• The final output is written to a file path following a slug-based naming convention and must adhere strictly to the provided template structure.

When to use it

• Use when you need a formal, evidence-based comparison between SAM and a target methodology to identify overlap/divergence, weaknesses in SAM, and implementation pairing opportunities. Use only when you have verifiable references read during this session and you will produce a document with 0 to 6 sections, fully completed.

What it can touch

• Tools indicated: claude-code is listed as a declared tool in the skill metadata.
• It relies on loading and reading local files (e.g., SAM document) and target methodology documentation, and on using a specified internal template file structure.

Caveats

• All comparisons must be based on verifiable reference material read during this session; it cannot rely on training data.
• Filenames and tokens from SAM docs are treated as examples, with taxonomy mapped to semantic tokens; direct canonical identifiers are avoided in interpretation.
• The output must use relative or markdown-linkable paths when referencing repository files and include exact citations with dates. It must follow the 6-section template and provide an evidence log for every claim.