The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download a CLI tool installer from the vendor's official domain (hindsight.vectorize.io) during the setup phase.
[REMOTE_CODE_EXECUTION]: The setup instructions include a pattern where a remote shell script is downloaded and piped directly to bash (curl -fsSL https://hindsight.vectorize.io/get-cli | bash). While this originates from the vendor's own infrastructure, it is a high-privilege execution pattern.
[COMMAND_EXECUTION]: The skill relies on executing the hindsight command-line utility for all core functions, including retain, recall, and reflect operations.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests and processes untrusted data from external sources.
Ingestion points: Raw conversation transcripts, session notes, and project logs are passed to the memory retain command in SKILL.md.
Boundary markers: Commands use standard shell quoting for input, but no specific semantic delimiters or instructions are provided to the agent to ignore embedded commands within the content being stored.
Capability inventory: The skill executes the hindsight CLI tool via subprocess calls and performs network requests to the vendor's API.
Sanitization: There is no evidence of local sanitization of the input strings before they are passed to the CLI tool or the Hindsight Cloud server.

hindsight-cloud