BB RAG
Designs, tunes, and debugs retrieval-augmented generation (RAG) pipelines: chunking, embeddings, hybrid retrieval, reranking, and grounded answers. Use when a system returns the wrong passages, misses a document that is indexed, cites nothing, or hallucinates over good context; when choosing a vector store, an embedding model, a chunk size, or a reranker; when similarity scores collapse after a model swap; when a metadata filter empties the result set; when answers ignore mid-context facts; when follow-up questions retrieve the wrong thing; when indexing PDFs, scanned pages, tables, code, or transcripts; when GDPR erasure, tenant isolation, or prompt injection from indexed documents is the problem; or when per-query cost or p95 latency has to come down. Covers reindex migrations, corpus freshness, evaluation sets, and agentic and graph retrieval. Not for splitter internals (`rag-chunking`), scoring rubrics (`rag-evaluation`), or LangChain APIs (`langchain`).
Designs, tunes, and debugs retrieval-augmented generation (RAG) pipelines: chunking, embeddings, hybrid retrieval, reranking, and grounded answers.
As a process B 72/100 · Nearly there — weak spots: inputs and preconditions, running it twice
How to improve
- The SKILL.md body is over 5,000 tokens: move reference detail into references/ and load it when needed.
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 1
✓ No critical or high findings
Medium and low: 1
-
low Risky intent
intent-offensive-securitysecurity.md:125Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| Permission filter derived from anything the user can type | Privilege escalation by prompt | Move derivation server-side, from the session, today |
Files scanned: 20. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- warning
name-formatname should be kebab-case (lowercase letters, digits, hyphens) - warning
body-longSKILL.md body ≈ 7390 tokens (recommended < 5000); move details to references/ - note
frontmatter-keyunknown frontmatter key "slug" - note
frontmatter-keyunknown frontmatter key "homepage" - note
frontmatter-keyunknown frontmatter key "changelog"
Process rating: all ten parameters 72/100
- 0Inputs and preconditions. Does not say what the process needs to start
- 30Running it twice. 17 mutating operations with no state check
- 60Tools and files. Uses tools (web) that frontmatter does not declare
- 60Result and completion. Output format stated, no completion criterion
- 70Execution cost. Instruction body is 7390 tokens
- 100Steps. 43 steps
- 100When it triggers. States when to use and when not to
- 100Failures and branches. 2 branches, has a failure section
- 100Consistency. Name and required fields are in place
- 100Progress reporting. Reports progress
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
- low 13 top-level sections: this looks like several domains in one skill
- low The skill ranks results itself: that belongs to the system behind the tool, not the model
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +5Description has no quoted example phrases that should trigger the skill
- +3Description length 972: 120–800 characters recommended
- +4No input/output examples
- +1No license
- +2Single-language instructions
- +4Description says when NOT to use the skill
- +4Structure: 13 headings
- +3Step-by-step instructions: 43 items
- +3Output format is stated explicitly
Quality base 70; lint remarks subtract, signals add up to 100. Result: 66.