The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to access and process sensitive personal information stored in Google Chrome's local database files. These files contain a comprehensive record of the user's browsing history and bookmarks.
Evidence: scripts/extract_chrome_activity.py accesses browser data paths such as ~/Library/Application Support/Google/Chrome/Default/History (macOS), %LOCALAPPDATA%\Google\Chrome\User Data\Default\History (Windows), and ~/.config/google-chrome/Default/History (Linux).
Context: Although this is the intended purpose of the skill, the extraction of browser history represents a high level of data exposure. The skill attempts to mitigate risk by using read-only database connections, processing data locally without network transmission, and setting restrictive file permissions (0o600) on temporary output files.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external, untrusted sources (web page titles and URLs in browsing history) which could contain malicious instructions meant to influence the AI agent's behavior.
Ingestion points: scripts/extract_chrome_activity.py reads data into listful-chrome-activity.json which the agent then analyzes.
Boundary markers: The instructions in SKILL.md do not define specific delimiters or instructions for the agent to ignore potential commands embedded within the extracted browser data.
Capability inventory: The agent has the capability to execute shell commands and write files to the local system based on the analysis of this untrusted data.
Sanitization: The extraction script normalizes URLs to remove credentials but does not sanitize page titles, which are a primary vector for indirect prompt injection.
[COMMAND_EXECUTION]: The skill workflow requires the agent to execute local Python scripts to perform data extraction and schema validation.
Evidence: SKILL.md explicitly instructs the agent to run python3 scripts/extract_chrome_activity.py and python3 scripts/validate_listful_bundle.py.

listful-personal-home