The Agent Skills Directory

[COMMAND_EXECUTION]: The stock-log.sh script is vulnerable to command injection in the show_latest function. It performs arithmetic expansion $((...)) on the unvalidated $n argument provided via the CLI, which allows for the execution of subshells or arbitrary commands via payloads like $(command).
[COMMAND_EXECUTION]: Multiple functions, including search_by_date, search_by_stock, and search_keyword, pass user-influenced variables directly into shell commands and subshells within double quotes. This allows for arbitrary command execution through shell expansion if an attacker provides a malicious keyword or date.
[DATA_EXFILTRATION]: The lack of input sanitization in the grep search functions facilitates argument injection. A malicious user could provide additional flags (e.g., -f) or extra file paths to manipulate the command into reading and displaying sensitive system files instead of the intended stock logs.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it reads data from stock-analysis-log.md, a file that aggregates external analysis data. Malicious instructions embedded in stock data could influence the agent's behavior when the log is retrieved.
Ingestion points: /root/.openclaw/workspace/memory/stock-analysis-log.md
Boundary markers: Absent in the script output.
Capability inventory: Shell script execution (grep, cat, tail, head).
Sanitization: Absent; the script outputs raw file content to the console.

stock-log-viewer