The Agent Skills Directory

[DYNAMIC_EXECUTION]: The loadLckResults function in scripts/_lib.js dynamically imports the lck-analytics or lck-results package by resolving its path at runtime. The resolution logic includes checking the GLOBAL_NPM_ROOT environment variable and executing a shell command to find the global npm root, creating a risk of loading untrusted code if the environment or global packages are compromised.
[REMOTE_CODE_EXECUTION]: The SKILL.md and instruction.md files instruct the agent to execute remote code from the npm registry using npx -y @nomadamas/k-skill@0. This package provides the primary instructions, file listings, and updates for the skill, effectively handing over control of the agent's behavior to a remote package from an untrusted organization.
[COMMAND_EXECUTION]: The scripts/_lib.js file uses child_process.execFileSync to execute the npm root -g command. While this is used for environment detection, it demonstrates the skill's ability to execute shell commands to interact with the host system.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data, such as Oracle's Elixir style CSV files (scripts/sync-oracle.js) and live game JSON payloads (scripts/analyze-live-game.js). These ingestion points represent a surface for indirect prompt injection attacks if the data sources are untrusted.
Ingestion points: scripts/sync-oracle.js (CSV input), scripts/analyze-live-game.js (live stats JSON via --window and --details arguments).
Boundary markers: No explicit delimiters or instructions are used to separate ingested data from agent instructions.
Capability inventory: The skill can write files to the local reports/cache directories and execute shell commands (npm).
Sanitization: The scripts use standard JSON.parse and JSON.stringify but do not perform validation or sanitization of the natural language content within the data.

lck-analytics