The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute a local script python3 script/l3_core.py. This script is used to verify the state of the problem set, generate status reports, and rebuild the core data file (data/l3_core.json) when necessary.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from local repository files, which constitutes a potential attack surface.
Ingestion points: Reads problem verdicts and metadata from data/progress.txt, data/l3_core.json, and README.md.
Boundary markers: The skill instructions direct the agent to parse the log using specific helpers to ensure consistency, but does not specify explicit LLM-level delimiters for the data content.
Capability inventory: The skill has access to Read, Glob, Grep, and Bash tools, allowing it to read the filesystem and execute the local Python script.
Sanitization: The data parsing is delegated to a local Python script (script/l3_core.py) which implements the logic for extracting verdicts like 'ok' or 'again', reducing the risk of the LLM directly executing malicious instructions hidden in the text files.
[EXTERNAL_DOWNLOADS]: The skill contains links to GitHub blobs and other skill files within the yennanliu repository. These are vendor-owned resources used for navigation and documentation purposes.

l3-core