The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes a workflow where the agent ingests and relies on content from local files such as .terrain/agent/context.md and .terrain/knowledge/**/*.md to guide its actions.
Ingestion points: The agent is instructed to read architecture summaries, private domain knowledge, and structured metadata from multiple markdown files within the repository's .terrain/ directory.
Boundary markers: The instructions lack explicit delimiters or warnings to treat the content of these files as untrusted or to ignore any embedded instructions found within them.
Capability inventory: The skill explicitly mentions loading other skills like rtk-skill for shell operations (git, grep, tests) and codegraph-skill for impact analysis, providing a path for instructions in the ingested data to influence command execution.
Sanitization: There is no evidence of sanitization or filtering applied to the external content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands using both local binaries and remote package runners.
Evidence: It suggests using ~/.terrain/bin/terrain or bunx @terrain-ai/cli to check the freshness of architecture context.
Context: These commands are used to interface with the Terrain toolset and are consistent with the skill's stated purpose of managing project knowledge.

terrain-knowledge-skill