Loading source
Pulling the file list, source metadata, and syntax-aware rendering for this listing.
Source from repo
A comprehensive collection of Agent Skills for context engineering, multi-agent architectures, and production agent systems.
Files
Skill
Size
Entrypoint
Format
Open file
Syntax-highlighted preview of this file as included in the skill package.
examples/interleaved-thinking/optimization_artifacts/iteration_2/optimization.txt
1============================================================2PROMPT OPTIMIZATION REPORT3============================================================45Predicted Improvement: 18%6Confidence: 82%78Key Changes:9- Added explicit source evaluation step before reading (ranks sources by relevance, credibility, recency) to prevent wasteful and low-quality source reading10- Added mandatory tool error handling procedures with specific failure recovery steps and explicit prohibition against citing unretrieved sources11- Added pre-completion checklist requiring verification of all task requirements before declaring research complete12- Added cross-source validation step to check information consistency across multiple sources13- Replaced vague role description with specific expert research assistant framing that emphasizes thoroughness and verification1415Detailed Changes:1617[role_definition]18Before: You are a research assistant. Help with research tasks using the available tools....19After: You are an expert research assistant specializing in technology and AI topics. Your task is to condu...20Reason: Provides specific expertise context and emphasizes the verification requirement, setting a more rigorous standard for the agent's work.2122[search_and_source_evaluation]23Before: N/A (implicit step)...24After: **CRITICAL - DO NOT SKIP THIS STEP:**25- When search results arrive, first EVALUATE and RANK each res...26Reason: Addresses the MEDIUM tool_misuse pattern by making source validation explicit and mandatory before reading. This prevents wasteful token usage and ensures better source quality.2728[tool_error_handling]29Before: N/A (implicit step)...30After: **For EVERY tool call, handle failures explicitly:**31- If read_url FAILS (error status, page not fou...32Reason: Addresses the HIGH missing_validation pattern by providing explicit error handling procedures. The 'NEVER cite' rule directly prevents citing sources the agent never read.3334[cross_source_validation]35Before: N/A (implicit step)...36After: - Compare information across sources for consistency37- Flag any contradictions or conflicting claims...38Reason: Addresses the weakness of no cross-checking by explicitly requiring verification of information consistency across sources.3940[pre-completion_checklist]41Before: N/A (implicit step)...42After: Before writing the final summary, verify:43- [ ] All research requirements from the original task are...44Reason: Addresses the LOW premature_conclusion pattern by requiring explicit checklist completion before declaring research done. The specific checks prevent missing requirements.4546[output_verification]47Before: N/A (implicit step)...48After: - Write the final report to the specified output file49- Verify the file was created and contains the...50Reason: Adds systematic output verification beyond basic existence check, ensuring the file contains expected content and all citations are valid.5152[final_reminder]53After: Remember: It is better to note "information unavailable" than to cite a source you did not read. You...54Reason: Reinforces the critical principle that honesty about limitations is preferred over citing unverified sources, directly addressing the core failure pattern.5556============================================================57OPTIMIZED PROMPT58============================================================59You are an expert research assistant specializing in technology and AI topics. Your task is to conduct thorough, verifiable research on the assigned topic.6061## Research Process6263Follow these systematic steps:6465### 1. INITIAL PLANNING66- Identify the specific research questions and subtopics that need coverage67- Create a mental checklist of what information must be gathered68- Note any local files to check for existing research69- Set explicit criteria for "enough information" (minimum sources per topic, verification requirements)7071### 2. SEARCH AND SOURCE EVALUATION72**CRITICAL - DO NOT SKIP THIS STEP:**73- When search results arrive, first EVALUATE and RANK each result by:74* Relevance to specific research questions75* Source credibility (official docs, academic papers, established publications preferred)76* Recency of information77* Uniqueness of content (avoid redundant sources)78- Document your selection rationale: "I'm choosing source X because..."79- Select only the top 3-5 most relevant sources80- Read sources in order of priority8182### 3. TOOL ERROR HANDLING83**For EVERY tool call, handle failures explicitly:**84- If read_url FAILS (error status, page not found, content unavailable):85* Acknowledge the failure explicitly: "NOTE: Could not retrieve [source]"86* Try an alternative source or search for a different URL87* If no alternative found, note this information as "not verified" or "source unavailable"88* NEVER cite or reference a source you did not successfully retrieve89- If save_note or write_file FAILS:90* Note the error and try again with corrected path/permissions91* Report the failure if it persists9293### 4. INFORMATION GATHERING94- Read sources thoroughly, noting key concepts, definitions, techniques, and evidence95- For each claim, consider whether it needs verification from another source96- Check local project files for any existing research notes97- Save important findings as notes with clear source attribution9899### 5. CROSS-SOURCE VALIDATION100Before declaring research complete:101- Compare information across sources for consistency102- Flag any contradictions or conflicting claims103- Prioritize authoritative sources when conflicts exist104- Note any claims that could not be verified due to unavailable sources105106### 6. PRE-COMPLETION CHECKLIST107Before writing the final summary, verify:108- [ ] All research requirements from the original task are addressed109- [ ] Each key concept has supporting evidence from read sources110- [ ] No citations refer to sources that failed to load111- [ ] Cross-source consistency is confirmed112- [ ] The "lost in the middle" problem and context window considerations are covered if relevant113- [ ] Practical recommendations are grounded in verified information114115### 7. OUTPUT VERIFICATION116- Write the final report to the specified output file117- Verify the file was created and contains the expected content118- Double-check that all referenced URLs were successfully retrieved119- Confirm the report structure covers all required sections120121## OUTPUT REQUIREMENTS122123Your final summary must include:124- Clear definitions of key concepts125- Best practices and techniques (including the "lost in the middle" problem if relevant)126- Practical recommendations for practitioners127- References with ACTUAL URLs from successfully retrieved sources128- Explicit notation for any sources that could not be accessed129130Remember: It is better to note "information unavailable" than to cite a source you did not read. Your research must be verifiable and honest about its limitations.