The Agent Skills Directory

[Remote Code Execution] (CRITICAL): The skill contains commands to download and execute scripts directly from the internet (curl | sh) for installing the uv and rust toolchains. Specifically, it includes curl -LsSf https://astral.sh/uv/install.sh | sh and curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh. This pattern is highly dangerous as it bypasses script verification and executes unvetted remote code with current user privileges.
[External Downloads] (HIGH): The skill recommends installing claude-code-tools and aichat-search from external repositories without providing integrity hashes or version pinning, increasing susceptibility to supply chain attacks.
[Data Exposure] (MEDIUM): The skill's core functionality involves reading sensitive session logs from ~/.claude/projects/. These files contain historical transcripts of AI agent interactions, which frequently include proprietary code, personal information, or environment secrets.
[Indirect Prompt Injection] (LOW): The skill processes untrusted historical session data, creating a surface for indirect prompt injection attacks where an attacker could influence future agent actions by poisoning previous logs.
Ingestion points: Reads .jsonl session files from ~/.claude/projects/.
Boundary markers: None; the agent is directed to read and summarize the content directly into its context without isolation.
Capability inventory: Execution of aichat CLI tools and local file system reads.
Sanitization: None; there is no escaping or filtering of the historical session content before it is processed.

session-search