The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it fetches unvetted company descriptions and financial summaries from Yahoo Finance and SEC EDGAR and interpolates them directly into LLM prompts.
Ingestion points: fetch_data.sh retrieves the longBusinessSummary field from Yahoo Finance and company facts from SEC EDGAR.
Boundary markers: Absent; external data is placed under a "Raw Data" header in run-framework.sh without delimiters or instructions to ignore embedded commands.
Capability inventory: The skill can write files to the assets/ directory, execute local bash scripts, and perform network requests via curl (as seen in fetch_data.sh).
Sanitization: No sanitization or filtering is applied to the text content retrieved from external APIs before it is sent to the LLM.
[COMMAND_EXECUTION]: The skill relies on the execution of multiple local bash scripts (analyze-pipeline.sh, fetch_data.sh, run-framework.sh) to orchestrate its 8-stage analysis pipeline.
[EXTERNAL_DOWNLOADS]: Fetches structured financial data from well-known and trusted services, including Yahoo Finance (query2.finance.yahoo.com), SEC EDGAR (data.sec.gov), and Alpha Vantage (alphavantage.co).
[CREDENTIALS_UNSAFE]: The script fetch_data.sh accesses the user's Alpha Vantage API key from the OpenClaw configuration file located at ~/.openclaw/agents/main/agent/auth-profiles.json. While this is a documented feature of the skill, it involves reading sensitive credential files.

company-analyzer