Research domain best practices and failure modes for agent skills.
npx skills add https://github.com/dvy1987/agent-loom --skill research-skill-dvy1987
Please help me install this Agent Skill. Skill: research-skill Source: https://github.com/dvy1987/agent-loom/tree/main/.agents/skills/research-skill Command: npx skills add https://github.com/dvy1987/agent-loom --skill research-skill-dvy1987
This Skill identifies up-to-date, non-obvious domain knowledge, failure modes, and existing skill patterns that a general LLM might miss so that new or improved agent skills are accurate, evidence-backed, and safe to adopt.
Research the domain sprint-retrospective and deliver a structured findings report listing sources, gotchas, workflow patterns, failure modes, and existing skills.
High-intent search queries and answers about installing and using this skill.
Researching domain-specific failure modes involves querying academic papers, practitioner blogs, and GitHub skill repositories in parallel to surface high-signal findings that a general LLM might miss. The process classifies findings into structured taxonomies like GOTCHA and FAILURE_MODE to ensure new agent skills are evidence-backed and safe to adopt.
Finding existing skill patterns and gotchas requires parallel multi-source searches across GitHub skill repos and academic papers like arXiv. This approach surfaces up-to-date, non-obvious domain knowledge and existing patterns that a general LLM might miss, which is then classified into a structured findings report.
Securely ingesting external SKILL.md files mandates invoking local secure-* scanners before consuming the content. This security-gated ingestion process filters and blocklists unsafe sources, ensuring that only verified, safe patterns are integrated into the skill-authoring and validation workflows.
A literature review can classify domain workflows into structured taxonomies such as GOTCHA, WORKFLOW, FAILURE_MODE, EXISTING_PATTERN, and BACKGROUND. This structured reporting is consumable by creation tools to improve the accuracy and safety of new agent skills.
A general LLM does miss non-obvious domain knowledge, failure modes, and existing skill patterns. Researching these domains through parallel source searches and security-gated ingestion ensures new or improved agent skills are accurate, evidence-backed, and safe to adopt.
plurigridQuery local ~/.topos notes and academic MCPs in parallel for multi-source research.
GovTechSGStandardize authoring and organizing AI agent skills with frontmatter and folder structure.
diarworldGenerate Agent Skills packages with automated security scanning.
falconntQuery web sources to verify factual claims with source attribution.
