git:20260531.39e76ac to git:20260905.c68a71a

27 added, 69 removed. Audit A to A.

---

name: websearch

- description: "Power search via Exa MCP. Modes: quick, deep research, code, docs, debug, news, compare. Use when searching the web, finding docs, debugging errors, or researching any topic."

+ description: Search the web for evidence using the available Exa tools. Prefer concise results and stop when the question is answered.

---

- <objective>

- Execute web searches via Exa MCP with intent-based routing, multi-pass deep research, and structured output with inline citations.

- </objective>

- <parameters>

- **Query**: free-text after the skill name

- **Modes** (mutually exclusive):

- - *(none)*: quick search, single pass, `web_search_exa`

- - `--deep`: multi-pass deep research with gap-driven iteration (3-5 sub-queries, max 3 passes)

- - `--code`: code examples, API docs, library usage via `get_code_context_exa`

- - `--docs <lib>`: crawl official documentation for a specific library

- - `--debug`: paste error message, find solutions (Stack Overflow, GitHub issues)

- - `--news`: recent news, category:news filter

- - `--compare`: compare 2+ options side by side via parallel queries

- - `--research`: academic papers, category:research paper

- - `--similar <url>`: find pages similar to a given URL

- **Filters**:

- - `--after <date>`: results published after date (ISO 8601 or natural: "last week")

- - `--before <date>`: results published before date

- - `--domain <d>`: restrict to domain(s), comma-separated

- - `--exclude <d>`: exclude domain(s)

- - `--fresh`: force live crawl (maxAgeHours=0)

- - `--locale <CC>`: country code for localized results

- - `-n <num>`: number of results (default: 5, max: 20)

- **Output**:

- - `--save <file>`: save report to markdown file

- - `--json`: structured JSON output

- - `--full`: full text instead of highlights (more tokens)

+ # Web search

- **Help**:

- - `--info`: display detailed tutorial with all modes, flags, and examples

- </parameters>

+ Use the live Exa tool schemas, not remembered parameter names. Quick lookup:

+ one focused search, then answer with citations. Documentation: find the

+ official page and crawl the relevant section only. Code/debug: use code

+ context, then broaden only when the first results do not answer the issue.

+ News: preserve publication dates. Compare: search the actual alternatives.

+ Research/similar: use supported search parameters; never invent an enum value.

- <state_variables>

- - {mode}: string (detected search mode: quick|deep|code|docs|debug|news|compare|research|similar)

- - {query}: string (cleaned search query)

- - {filters}: object (parsed Exa filters: dates, domains, locale, numResults)

- - {output_opts}: object (save path, json flag, full text flag)

- - {lib_name}: string (library name if --docs mode)

- - {similar_url}: string (URL if --similar mode)

- - {results}: array (accumulated search results across passes)

- - {pass_count}: number (current iteration for deep mode)

- - {gaps}: array (identified knowledge gaps for deep iteration)

- </state_variables>

+ Honor --domain/--exclude, --after/--before, -n, --fresh and requested output

+ format when the tool supports them; explain unsupported filters. Prefer

+ short relevant excerpts, not full pages. Crawl a full passage when needed

+ to verify a key claim. Do not paste raw response metadata twice.

- <entry_point>

- Load `steps/step-00-triage.md`

- </entry_point>

+ --deep permits targeted follow-ups for named remaining gaps, up to three

+ passes including the initial search. Stop sooner when answered, saturated or

+ no new useful evidence appears. No fixed quota of queries or sources. Run

+ independent tool calls together without launching agents for simple lookups.

- <step_files>

- | Step | File | Purpose |

- |------|------|---------|

- | 00 | steps/step-00-triage.md | Parse flags, classify intent, build Exa params |

- | 01 | steps/step-01-search.md | Execute search strategy per mode |

- | 02 | steps/step-02-deep.md | Deep iteration: evaluate, find gaps, re-query |

- | 03 | steps/step-03-report.md | Synthesize, cite, format, optionally save |

- </step_files>

+ Cite each source that supports a claim; distinguish observed facts, inference

+ and uncertainty. Keep dates honest. Write a report only with --save or an

+ explicit request. --json returns structured results. --info uses the existing

+ references/info-text.md.

- <references>

- | File | Content |

- |------|---------|

- | references/domain-presets.md | Curated domain lists per ecosystem/intent |

- | references/query-patterns.md | Optimal Exa query formulation per search type |

- | references/info-text.md | Full tutorial displayed by --info flag |

- </references>

+ ## Execution policy

- <critical>

- ALL web access via Exa MCP only. NEVER use WebSearch, WebFetch, or Context7.

- Tools: mcp__exa__web_search_exa, mcp__exa__get_code_context_exa, mcp__exa__crawling_exa, mcp__exa__web_search_advanced_exa.

- </critical>

+ Work solo. Ask before any subagent or reviewer, even in auto mode. Explain

+ the independent scope and expected benefit first. No hidden advisor, nested

+ delegation, model retuning, repeated successful checks, or progress spam.

+ Use existing context before asking questions. Stop when the requested result

+ is delivered. User stops and scope changes override pending steps.