The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted research data (customer voice, transcripts, notes, and desk research) to populate structured job-to-be-done frameworks. Maliciously crafted input in these sources could potentially influence the agent's behavior during the extraction (Stage 1) or auditing (Stage 3) phases.
Ingestion points: The skill reads raw source files, links, and pasted text during the extraction process.
Boundary markers: There are no explicit instructions or delimiters defined to prevent the agent from following instructions embedded within the research data.
Capability inventory: The skill has the ability to write to the file system (work/ directory) and execute local Python scripts that can invoke the system browser.
Sanitization: While the final HTML output is escaped for safety, the skill does not appear to sanitize the raw research data before it is processed by the LLM.
[COMMAND_EXECUTION]: The skill executes local Python scripts and utilizes system-level commands (open on macOS, os.startfile on Windows) to automatically display generated HTML files.
Evidence: The script scripts/render_html.py uses subprocess.run(['open', p]) and os.startfile(p) to open the resulting report file in the browser if the --open flag is provided.

jtbd-research-to-html