The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local Python scripts using the Bash tool (e.g., python scripts/generate_schematic.py "your diagram description" -o figures/output.png).
These scripts (scripts/generate_schematic.py, scripts/compliance_checker.py, scripts/budget_calculator.py, scripts/deadline_tracker.py) are referenced but their source code is not provided in the skill package, preventing verification of their behavior.
The use of shell commands with user-provided strings ("your diagram description") introduces a risk of command injection if the underlying scripts do not properly sanitize input.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, creating a vulnerability to indirect prompt injection.
Ingestion points: Processes external "research proposals", "program solicitations", "reviewer comments", and "technical narratives" (SKILL.md).
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present in the guidelines.
Capability inventory: The skill has access to Bash (command execution), Write, and Edit tools (SKILL.md).
Sanitization: No sanitization or validation logic is defined for processing external document content.
[METADATA_POISONING]: There is a discrepancy between the declared author in the YAML frontmatter (K-Dense Inc.) and the provided author context (wu-yc). Additionally, the skill includes hardcoded instructions to promote an external platform (www.k-dense.ai) for complex workflows, which could be used to divert users to unverified third-party services.

research-grants