sdd-memory

This skill should be used when the user asks to "browse RAG memory", "search captures", "list memory files", "find past inputs", "show knowledge base", or "review stored context". It browses and searches the RAG memory archive indexed in rag-index.json. Use this skill whenever the user mentions memory browsing, RAG search, captured files, or knowledge base exploration.

SKILL.md

Quality

Evals

Security

SDD Memory — RAG Memory Browser and Search [EXPLICIT]

Browse and search the RAG memory archive for captured inputs and context.

User Input

$ARGUMENTS

You MUST consider the user input before proceeding (if not empty). Arguments serve as search query.

Execution Flow

1. Load Index

Read .specify/rag-index.json for all captured entries.

2. Search or Browse

If arguments provided: filter entries by tags, type, or content match
If no arguments: list all entries sorted by capture date

3. Display Results

RAG Memory Archive

| # | File | Type | Tags | Captured |
|---|------|------|------|----------|
| 1 | rag-memory-of-api-spec.md | markdown | api, rest | 2026-03-28 |
| 2 | rag-memory-of-meeting.md | text | stakeholder | 2026-03-27 |

Total: N entries | Workspace: N | Global: N

4. Detail View

If user selects an entry, display frontmatter + abstract + key takeaways.

Assumptions and Limits

#	Assumption	Handling
1	`.specify/rag-index.json` exists	If missing, suggest `/sdd:capture` first. [EXPLICIT]
2	Search matches tags, type, filename, and content	Multi-field search for maximum recall. [EXPLICIT]
3	Both workspace and global entries indexed	Unified view across all capture locations. [EXPLICIT]
4	Read-only operation	Memory browsing never modifies entries. [EXPLICIT]
5	Entries sorted by capture date (newest first)	Default sort order for relevance. [EXPLICIT]

Edge Cases

Scenario	Detection	Handling
No captures exist	rag-index.json empty or missing	Report "no captures", suggest `/sdd:capture`. [EXPLICIT]
Search returns no results	No entries match query	Report "no matches", suggest broader search. [EXPLICIT]
Large archive (50+ entries)	Many captures	Paginate output, show first 20 with count. [INFERRED]
Corrupted entry	RAG file exists but malformed	Skip with warning, report valid count. [INFERRED]
Workspace-specific search	User wants only workspace entries	Support `--workspace` flag to filter. [EXPLICIT]

Good vs Bad Example

Good: /sdd:memory api

RAG Memory — Search: "api"

| # | File | Type | Tags | Captured |
|---|------|------|------|----------|
| 1 | rag-memory-of-api-spec.md | markdown | api, rest, auth | 2026-03-28 |
| 2 | rag-memory-of-api-notes.md | text | api, meeting | 2026-03-25 |

2 results found (of 8 total entries)

Bad: Memory listing without metadata

x Plain file list without tags or dates
x No search capability
x No workspace/global distinction

Why: Memory browser must provide searchable, tagged, dated entries with workspace awareness. [EXPLICIT]

Validation Gate

Before marking memory browse as complete, verify: [EXPLICIT]

V1: rag-index.json loaded or error reported
V2: Entries displayed with metadata (type, tags, date)
V3: Search filtered correctly if query provided
V4: Entry count reported
V5: Workspace vs global distinction shown

Repository: JaviMontano/mao-sdd
Path: .claude/skills/sdd-memory/SKILL.md
Commit: 3eafb9b

Last updated: 3 days ago
First committed: 6 months ago

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.