The Agent Skills Directory

[DATA_EXFILTRATION]: The skill requires access to sensitive browser profile directories (e.g., ~/Library/Application Support/Google/Chrome/, ~/.mozilla/firefox/) to extract session cookies for authentication. Access to these paths exposes sensitive user credentials and session data.\n- [COMMAND_EXECUTION]: The skill instructs the agent to execute multiple shell commands, including the peep CLI, file system operations (mkdir, cat), and data processing pipelines involving curl, jq, and xargs for media management.\n- [EXTERNAL_DOWNLOADS]: The skill requires cloning its source code from its GitHub repository (github.com/devskale/peep.git) and includes instructions for using curl to download media content from external URLs provided by the Twitter API.\n- [PROMPT_INJECTION]: The skill fetches content from external sources (tweets, search results, and news) which creates a surface for indirect prompt injection. Malicious data from these sources could influence the agent's behavior.\n
Ingestion points: Data enters the context via commands like peep read, peep search, peep home, and peep news.\n
Boundary markers: Absent; there are no instructions to the agent to isolate or ignore embedded commands in the fetched data.\n
Capability inventory: The skill provides write-access capabilities including peep tweet, peep reply, peep follow, and peep unfollow.\n
Sanitization: Absent; the skill does not perform validation or filtering on fetched text before it is presented to the agent.

peep