The Agent Skills Directory

[METADATA_POISONING]: The author field in the YAML frontmatter is set to "Claude Code", which is the name of an official tool developed by Anthropic. This attribution is deceptive as the skill is hosted and maintained by a third-party vendor ("inbharatai"), potentially misleading users about the skill's origin and safety vetting.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and act upon untrusted external data.
Ingestion points: The skill ingests content from external PDF and HTML forms as described in the overview and instructions (SKILL.md).
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are provided to the agent to prevent it from following instructions hidden within form fields or document metadata.
Capability inventory: The skill utilizes file manipulation capabilities through the pypdf2 library to generate output documents.
Sanitization: While the guidelines mention input validation, there is no specific mechanism or instruction to sanitize prompt-sensitive content from the processed files.

form-filler