The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/read_sheet.py saves sensitive Google OAuth2 tokens to a local file token.json. Storing credentials in predictable plaintext files on the local filesystem increases the risk of exposure if the environment is shared or compromised.\n- [COMMAND_EXECUTION]: The SKILL.md instructions direct the agent to execute Python scripts via the bash shell using heredocs to pass dynamically generated JSON input. While the use of quoted heredocs (<<'EOF') prevents shell interpolation, the script still executes logic based on LLM-generated data.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it extracts strategic content from untrusted sources, such as sales call transcripts, without using boundary markers or sanitization.\n
Ingestion points: The call_transcript input field in SKILL.md accepts raw external data that is processed to generate proposal content.\n
Boundary markers: Absent. The instructions do not define delimiters (e.g., XML tags) to isolate the transcript data from the agent's internal reasoning.\n
Capability inventory: The skill possesses network access to the PandaDoc and Google APIs via requests and gspread, and has write access to the filesystem for temporary data and credential storage.\n
Sanitization: Absent. The skill extracts problems and benefits directly from user-provided transcripts and interpolates them into formal document tokens without validation.

create-proposal