The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions include shell command templates (python scripts/dfs_client.py ... and python scripts/report_publish.py ...) that interpolate user-provided inputs such as <target query> and <report>.md. There are no instructions for sanitizing these inputs, which could lead to command injection if a malicious user provides specially crafted values containing shell metacharacters.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process text from external URLs or user pastes, creating a surface for indirect prompt injection where malicious instructions in the source text could hijack the agent's behavior.
Ingestion points: Untrusted text provided by the user or fetched from remote URLs via the webfetch tool.
Boundary markers: Absent. No delimiters or instructions are provided to the agent to distinguish between the content to be analyzed and potential instructions within that content.
Capability inventory: The skill utilizes shell execution for local scripts (dfs_client.py, report_publish.py), writes to the local file system (NLP-PASS-...md), and performs network operations.
Sanitization: Absent. No sanitization or validation of the input text or the interpolated shell arguments is specified before processing.

nlp-optimization