The Agent Skills Directory

[PROMPT_INJECTION]: User-provided search queries are directly interpolated into the system prompt sent to the xAI API in scripts/search.js. An attacker could provide a query designed to override the search assistant's instructions.
Evidence: input: `${systemPrompt}\n\nSearch X/Twitter for: ${options.query}\n\nReturn actual tweets with:...`
[DATA_EXFILTRATION]: The skill accesses the local configuration file ~/.clawdbot/clawdbot.json to retrieve API keys. While this is used for the skill's primary purpose, it demonstrates access to potentially sensitive local data.
Evidence: const configPath = path.join(process.env.HOME, '.clawdbot', 'clawdbot.json');
[COMMAND_EXECUTION]: The skill is designed to execute local Node.js scripts to perform searches, as defined in package.json and SKILL.md.
Evidence: "bin": { "search-x": "./scripts/search.js" } in package.json.
[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection (Category 8) as it processes untrusted data from X/Twitter which may contain malicious instructions.
Ingestion points: The skill fetches tweet content and metadata via the xAI API in scripts/search.js.
Boundary markers: No delimiters or "ignore embedded instructions" warnings are used when presenting the search results to the parent agent.
Capability inventory: The skill executes local commands via Node.js as defined in SKILL.md.
Sanitization: No sanitization or escaping is performed on the tweet content before it is output.

search-x