agenticnotetaking/skills/gid

/eidos:gid

Bidirectional translation between prose and GID (General Idea Distillation) notation. Two modes: compress (prose → GID) and decompress (GID → prose).

Usage

/eidos:gid compress <source>    # prose → GID
/eidos:gid write <source>       # alias for compress

/eidos:gid decompress <source>  # GID → prose
/eidos:gid read <source>        # alias for decompress

Source can be: a file path, multiple file paths, a URL, or inline text in the conversation.

Instructions

Reference

Before processing, internalise the full GID symbol table and conventions from [[spec - gid skill - semantic symbols for compressed notation]]. The full symbol set applies here, not just the eidos subset.

Key principles:

• Each symbol extracts the goal/aim/kind of an idea, removing the words that would otherwise express it
• Tree nesting groups related ideas under general parents — deeper = more specific
• The first character(s) of each line signal the kind of content
• Compression is lossy by design — favour scan-speed over completeness
• Implicit context during writing may not be present during reading — balance compression with recoverability

Compress Mode

1. Read the source — file(s), URL content, or inline text
2. Identify structure — find the main topics, subtopics, claims, questions, examples, etc.
3. Map to GID symbols — for each unit of meaning, identify which GID symbol captures its kind:
   • Is it a claim? → ^
   • A question? → ?
   • An example? → §
   • A definition? → :=
   • Important? → ! or !!
   • A goal? → 0
   • A problem? → S
   • A solution? → =
   • A follow/consequence? → -> or =>
   • A reason? → <-
   • A counter-point? → <
   • Pros/cons? → + / -
   • A source reference? → //
   • A todo? → [ ] (with variants [!], [?], [x], etc.)
   • A note? → > or . for side-notes
   • An idea? → °
   • A condition? → x
4. Build the tree — nest related points under their parents, using indentation (2 spaces per level)
5. Compress inline — use {..} for inline branches, | to chain symbols, _ for continuation
6. Remove redundant words — the symbol already carries the meaning; strip linguistic scaffolding like "this is an example of", "the reason is", "it's important that"

Compression Guidelines

• Top level = main topics or goals (0, >, ^)
• Second level = supporting points, explanations, conditions
• Third level+ = details, examples, sources, todos
• Chain symbols with | when a point has multiple aspects: //|@ Peters (source + person)
• Use {..} for brief inline asides that don't warrant their own line
• Use _ when a point continues on the next line
• Tags {TAG} for categorisation, abbreviations [AB]~ for recurring terms
• Don't over-compress: if a point would be unrecoverable without its original context, keep more words

Output

Write the result to <source-basename>.gid (or <name>.gid for inline/URL input). If the source is a single file notes.md, output to notes.gid in the same directory. If the user specifies a different output path, use that.

Decompress Mode

1. Read the GID source — file or inline text
2. Parse the tree — follow indentation to reconstruct the hierarchy
3. Resolve symbols — expand each GID symbol back to natural language:
   • ^ → "[Author] claims that..." or "Claim: ..."
   • ? → "Question: ..." or "Is it true that...?"
   • § → "For example, ..."
   • := → "[Term] is defined as ..."
   • ! → "Important: ..."
   • 0 → "Goal: ..." or "The aim is ..."
   • S → "Problem: ..."
   • = → "Solution: ..."
   • -> → "This leads to ..."
   • => → "Therefore, ..." or "Conclusion: ..."
   • <- → "The reason is ..."
   • < → "On the other hand, ..." or "Alternatively, ..."
   • + → "Pro: ..." or "Advantage: ..."
   • - → "Con: ..." or "Disadvantage: ..."
   • // → "Source: ..."
   • [ ] variants → "TODO: ..." with appropriate status
   • > → (just the note text)
   • ° → "Idea: ..."
   • x → "Condition: ..." or "When ..."
   • @ → person name
   • d → date
   • v → location
4. Reconstruct prose — turn the tree back into flowing paragraphs or structured sections
   • Respect the hierarchy: top-level points become paragraphs or section headers
   • Nested points become sentences within those paragraphs
   • Inline branches {..} become parenthetical remarks or footnotes
   • Chained symbols | get expanded to their respective meanings
5. Preserve intent — decompress recovers meaning, not the original text verbatim. The output should be clear, readable prose that captures what the GID notes encode.

Output

Write the result to <source-basename>.md (or present inline if the source was inline). If the source is notes.gid, output to notes.md in the same directory. If the user specifies a different output path, use that.

Notes

• Compression quality depends on the source material — highly structured text compresses better than conversational prose
• Decompression is inherently interpretive — GID symbols are intentionally vague, so the expansion involves judgement calls
• Round-tripping (compress → decompress) will not reproduce the original text, but should preserve the core meaning
• When uncertain about a symbol mapping during compression, prefer the more general symbol and keep more words
• The .gid extension is a convention — the content is plain text with GID notation