| name | ess - External Source Search |
| description | Systematic procedure for searching external sources to find evidence, data, and information relevant to a question. |
| output | {"format":"prose"} |
External Source Search
Input: $ARGUMENTS
Overview
Structured approach to finding information from external sources. Defines what you're looking for, selects appropriate source types, executes a systematic search, evaluates what you find, and synthesizes results. Prevents both insufficient searching and unfocused browsing.
Steps
Step 0: Check Capabilities
Do you have web search tools available in this session? If yes, proceed normally. If no: state what you know from training data, flag it as unverified, and recommend specific search queries the user can run. Do not simulate a search process you cannot actually perform.
Step 1: Define the Search
- What specific question are you trying to answer?
- What kind of evidence would answer it? (data, studies, expert opinion, examples, documentation)
- What would a satisfying answer look like?
- What's the quality threshold? (rigorous research vs quick orientation)
- Time budget: how long is this search worth?
Step 2: Select Source Types
Choose based on what you need:
| Need | Source Types |
|---|
| Scientific evidence | Academic databases, journals, preprints |
| Current events/trends | News, industry publications, social media |
| Technical documentation | Official docs, wikis, Stack Overflow, GitHub |
| Market/business data | Industry reports, financial databases, company filings |
| Expert opinion | Books, talks, interviews, podcasts |
| Historical precedent | Archives, case studies, historical databases |
| Public data | Government databases, census, open data portals |
Step 3: Construct Search Queries
- Primary query: key terms from the question
- Synonyms and related terms
- Boolean operators where supported (AND, OR, NOT)
- Domain-specific terminology
- Plan 3-5 query variations to cast a wide net
Step 4: Execute Search
- Search each source type with each query variation
- For each result, record:
- Source (where found)
- Key claim or data point
- Relevance to original question (high/medium/low)
- Quality signal (peer-reviewed, reputable, unknown)
- Stop when: finding repetition (same info from multiple sources) OR time budget exhausted
Step 5: Evaluate Sources
For each relevant source:
- Credibility: Who produced this? What's their expertise? Conflicts of interest?
- Methodology: How was this information generated?
- Recency: When was this produced? Still current?
- Corroboration: Do other sources agree?
- Assign tier: A (high quality) / B (moderate) / C (low) / D (unreliable)
→ INVOKE: /src [source] for deep credibility evaluation if needed
Step 6: Synthesize
- What do the sources collectively say?
- Where do they agree? Disagree?
- What's the answer to the original question based on what was found?
- What's the confidence level? (depends on source quality and agreement)
- What wasn't found? (gaps in the search)
SEARCH RESULTS:
Question: [original question]
Sources searched: [N types, N queries]
Results: [N sources found, N relevant]
Key findings:
1. [finding] — Source: [ref] — Quality: [tier]
2. [finding] — Source: [ref] — Quality: [tier]
Answer: [synthesis]
Confidence: [high/medium/low]
Gaps: [what wasn't found]
When to Use
- Any question requiring external information
- Fact-checking claims
- Research for decisions or plans
- Finding precedents or examples
Verification
