The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized data access attempts were detected within the skill's instructions or metadata.
[COMMAND_EXECUTION]: The skill references local utility commands, such as curl for health checks on localhost:9090 and git remote for retrieving project identifiers. These are standard diagnostic and configuration operations relevant to the tool's functionality.
[PROMPT_INJECTION]: The skill facilitates the ingestion of project data and session history, creating a surface for indirect prompt injection.
Ingestion points: Data enters the context via semantic_search, repository_index, and memory retrieval tools.
Boundary markers: None explicitly defined in the markdown instructions.
Capability inventory: File reading, semantic indexing, and persistent memory recording.
Sanitization: The skill explicitly highlights that contextd v1.5+ performs strict input validation, including the rejection of directory traversal (../), filtering of shell injection characters in glob patterns, and scrubbing of secrets from conversation histories.

using-contextd