The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides Python code that uses subprocess.run to configure the mcporter tool. It dynamically sets configuration parameters and authorization headers using tokens retrieved during the authentication flow.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to vendor-controlled domains (api.xpoz.ai, mcp.xpoz.ai) using urllib.request and curl. These operations are used for client registration, trial token acquisition, and OAuth token exchange.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its core function of ingesting and analyzing untrusted social media content.
Ingestion points: Untrusted data enters the agent context via the getTwitterPostsByKeywords and getRedditPostsByKeywords tools as described in SKILL.md.
Boundary markers: Absent; the instructions do not implement delimiters or specific warnings to the agent to ignore instructions embedded within the social media posts.
Capability inventory: The skill environment allows for filesystem writes (~/.cache/xpoz-oauth/state.json), shell command execution via subprocess.run, and network access via Python libraries and CLI tools.
Sanitization: Absent; there is no evidence of content filtering, escaping, or validation applied to the external data before it is processed by the agent.
[CREDENTIALS_UNSAFE]: The skill handles sensitive tokens and API keys. It utilizes standard practices such as OAuth flows and environment variables for management. It stores temporary state in ~/.cache/xpoz-oauth/state.json, which is a local configuration directory.

security-osint