The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple external sources, including Reddit, X, and general web articles, to generate research reports and synthesized prompts.
Ingestion points: Data is collected through web_search, reddit_search (or the reddit-insights MCP), bird search (via Bird CLI), and web_fetch as described in both SKILL.md and README.md.
Boundary markers: While the skill provides structured Markdown templates for its output, it lacks explicit instructions or boundary markers to prevent the agent from following malicious instructions that might be embedded within the ingested social media posts or web content.
Capability inventory: The skill possesses the capability to perform network operations and generate new prompts based on aggregated data, which could be leveraged to execute actions requested by an attacker-controlled source.
Sanitization: There is no evidence of sanitization or filtering logic applied to the external content before it is processed by the model for synthesis.
[COMMAND_EXECUTION]: The skill relies on external CLI tools and shell-like execution patterns for its core functionality.
The setup verification process involves shell operations such as sourcing an environment file (source ~/.openclaw/credentials/bird.env) and executing the bird CLI tool.
Search queries are interpolated directly into shell commands (e.g., bird search "[topic]"), which could lead to command injection if the agent does not properly escape shell metacharacters in the user-provided topic.

last30days