The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits a vulnerability surface for indirect prompt injection as it aggregates data from untrusted external web sources to produce structured reports.
Ingestion points: The skill uses web.search and web_fetch tools to gather commercial reports, exhibition information, and transformation cases from the public web as specified in SKILL.md.
Boundary markers: There are no explicit instructions or delimiters defined in SKILL.md or the analysis framework to isolate or ignore potentially malicious instructions embedded in the external data.
Capability inventory: The skill performs file-writing operations, creating HTML reports via the files tool and Word documents using python.run with the python-docx library (as seen in scripts/generate_word_report.py).
Sanitization: No evidence of sanitization, escaping, or validation of the ingested web content was found in the provided scripts or instructions.

smart-construction-analysis