The Agent Skills Directory

[PROMPT_INJECTION]: The skill retrieves and processes untrusted data from the web using WebSearch and WebFetch, which creates a surface for indirect prompt injection attacks. Malicious instructions embedded in third-party websites could attempt to manipulate the agent's output or logic.
Ingestion points: External web content retrieved via search and fetch tools (documented in SKILL.md sections 3.2 and 4).
Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore' instructions when passing retrieved data to the LLM for drafting.
Capability inventory: The skill has the ability to read and write local configuration files, execute shell commands, and perform network requests.
Sanitization: No technical sanitization or filtering of the fetched web content is performed within the skill's instructions.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute wewrite sources add with arguments like --title and --claim that are populated with content directly from external web pages. This presents a command injection surface if the external content contains shell metacharacters and is not properly escaped before being executed in the shell.
[DATA_EXFILTRATION]: The skill initiates network requests to external domains via WebSearch and WebFetch to gather information. While intended for research, these operations involve communicating with non-whitelisted servers.

wewrite-write