The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill contains logic to automatically extract and cache Bilibili SESSDATA cookies from local Chrome, Edge, and Firefox profiles. Extracted cookies are stored in ~/.config/bbc-skill/cookie.json with restricted permissions (0600).
[PRIVILEGE_ESCALATION]: In bbc/cookie/chrome_macos.py, the skill uses the macOS security command to retrieve the 'Chrome Safe Storage' password from the system Keychain. This key is required to decrypt local browser cookies.
[COMMAND_EXECUTION]: The skill executes several system binaries via subprocess.run, including security for keychain access and openssl for AES decryption of browser data. Additionally, SKILL.md instructs the agent to execute git commands for update checks and repository pulls.
[EXTERNAL_DOWNLOADS]: The skill's instructions guide the agent to perform a git pull from the vendor's repository (https://github.com/Agents365-ai/bbc-skill) to update the skill's source code, which constitutes a remote code update mechanism.
[INDIRECT_PROMPT_INJECTION]: The skill fetches untrusted user-generated content (Bilibili comments) and flattens it into JSONL files for subsequent agent analysis.
Ingestion points: Fetches comments from api.bilibili.com (seen in bbc/api.py and bbc/fetch.py).
Boundary markers: The output is structured JSONL, but the message field contains raw, unsanitized comment text without explicit boundary markers or instructions to ignore embedded commands.
Capability inventory: Agents using this skill typically have capabilities like file system access and tool execution to process the resulting data.
Sanitization: The skill performs standard JSON encoding but does not sanitize or filter the content of comments for potential injection patterns.

bbc-skill