The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes subcommands of the autocontext CLI tool, which is a required prerequisite for interacting with the Autocontext platform.
[DATA_EXFILTRATION]: The skill transmits data to and from the Autocontext API and Pinecone vector databases. These are well-known vendor services associated with the skill's author, pinecone-io.
[CREDENTIALS_UNSAFE]: The skill manages authentication via the autocontext login command and mentions storing credentials in ~/.autocontext/autocontext.json. This is standard behavior for CLI-based authentication.
[PROMPT_INJECTION]: The skill supports workflows for ingesting external documents and performing retrieval tasks, which presents an attack surface for indirect prompt injection. * Ingestion points: Data is ingested into contexts using the autocontext task create --workflow ingest command as documented in SKILL.md. * Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded content within the ingested files. * Capability inventory: Tool access is restricted to the autocontext binary via the allowed-tools policy in the YAML frontmatter. * Sanitization: The skill instructions do not implement or specify sanitization logic for data processed during ingestion or retrieval tasks.

autocontext