The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the vendor-specific clay CLI tool to verify the environment and list available resources.
Evidence: Shell commands in SKILL.md Step 0: clay whoami; echo "exit_code=$?", clay routines list --limit 100, clay routines get <routineId>, and clay workflows actions list.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests untrusted data from external company websites and potentially from local 'answer sheets'.
Ingestion points: External company domains (homepage, /about, /product) as described in SKILL.md Step 2 and references/research-recipe.md Phase 1; local 'answer sheets' mentioned in SKILL.md Declared inputs.
Boundary markers: Absent; there are no specific instructions for the agent to ignore or delimit embedded instructions within the scraped web content.
Capability inventory: Shell command execution via clay CLI (SKILL.md Step 0) and local file write capabilities to save research answers (SKILL.md Section: Declared inputs).
Sanitization: Absent; the skill relies on the agent to summarize and classify the data without explicit sanitization or filtering of the raw scraped text.
[EXTERNAL_DOWNLOADS]: The skill fetches data from external company websites via scraping and calls vendor-managed enrichment and news APIs.
Evidence: Mentions of scrape-website, managed 'Enrich Company' function, and 'Company News' catalog actions in SKILL.md and references/research-recipe.md.

company-research-brief