The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute various shell commands including ls, grep, and rm to inspect the project structure and manage the .claude/session-context.toon cache file.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection by ingesting data from untrusted local files.
Ingestion points: Reads package.json, scans directories like src/components/, and loads configuration/markdown files from .claude/project-contexts/.
Boundary markers: None identified; external content is parsed directly to generate the session context.
Capability inventory: Uses Read, Write, Grep, Glob, and Bash tools, providing a wide range of actions that could be influenced by injected instructions.
Sanitization: No evidence of sanitization or input validation for the content read from the project files.
[DYNAMIC_EXECUTION]: The skill triggers the execution of a local shell script (bash scripts/context-compress.sh) to perform manual context generation, which is a form of dynamic script execution based on local files.

project-context-loader