The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the ccl_chromium_reader package directly from a remote GitHub repository (https://github.com/cclgroupltd/ccl_chrome_indexeddb.git) using the uv pip install command.
[REMOTE_CODE_EXECUTION]: By installing unversioned code from a third-party GitHub repository, the skill introduces a risk of remote code execution if the source repository is compromised or contains malicious logic.
[DATA_EXFILTRATION]: The skill provides instructions and scripts to access and read sensitive browser storage files located in system directories such as %APPDATA%, ~/Library/Application Support/, and ~/.config/.
These files often contain session tokens, authentication cookies, cached messages, and other private user data from apps like Claude Desktop, Discord, and Slack.
The dump_localstorage.py and dump_indexeddb.py scripts extract this information and present it to the agent, which could be coerced into exfiltrating the data to an external endpoint.
[COMMAND_EXECUTION]: The skill uses shell commands to copy sensitive storage directories to temporary locations and modify file permissions (removing LOCK files) to facilitate reading potentially protected data.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. If a malicious application or website stores adversarial instructions within the Local Storage or IndexedDB of a targeted app, those instructions could be executed or obeyed by the agent when it parses and processes the storage records.
Ingestion points: scripts/dump_localstorage.py and scripts/dump_indexeddb.py read raw key-value pairs from LevelDB stores.
Capability inventory: The skill is granted Read, Write, and Bash tools.
Sanitization: The provided scripts perform no sanitization of the extracted data before outputting it to the agent's context.

leveldb-ops