The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute curl for API interactions and grep for searching local files. These operations are scoped and necessary for the core functionality of tracing artifact impact chains.
[EXTERNAL_DOWNLOADS]: In 'Service Mode', the skill communicates with an external graph engine via the $ENGRAM_URL environment variable. It sends impact analysis queries to this endpoint using curl. This behavior is transparently documented and refers to the vendor's own 'EngramGraph' project.
[DATA_EXFILTRATION]: The skill processes local Markdown files to reconstruct relationship chains in 'Degraded Mode'. While this presents an indirect prompt injection surface (Category 8) where malicious content in documents could attempt to influence the agent, the skill uses standard file reading and grep patterns to extract structured data (IDs and frontmatter), which is a common and acceptable pattern for documentation analysis tools.
Ingestion points: Local Markdown files identified via grep or Glob.
Boundary markers: None explicitly defined in the instructions for separating file content from system instructions.
Capability inventory: Bash(curl), Read, Glob, Grep tools are available.
Sanitization: The skill relies on standard string extraction (regex) for IDs but does not explicitly sanitize the text content before presentation.

knowledge-graph