The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands via uv run to perform searches, logins, and cart management. It also utilizes Playwright to control browser windows for store interactions and final checkout handoff.
[EXTERNAL_DOWNLOADS]: During setup, the skill runs playwright install chromium, which fetches browser binaries from official Playwright sources to facilitate the browser automation required for its tasks.
[CREDENTIALS_UNSAFE]: The skill manages authentication by storing live session cookies for TGP, Atomic Empire, TCGPlayer, and Mana Pool in the ~/.cache/mtg-skills/lgs-profiles/ directory. Although it specifies owner-only permissions (0700), these cookies are functionally equivalent to passwords for the target stores.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external card lists which provides a surface for indirect prompt injection.
Ingestion points: The skill reads user-provided card lists from files passed via the --input argument in SKILL.md.
Boundary markers: The instructions do not include specific boundary markers or delimitation requirements for the input card list content.
Capability inventory: The skill can execute subprocesses via uv run, perform network requests through Playwright, and write output files to the local system.
Sanitization: There is no documentation regarding the validation or escaping of input card names before they are used in CLI arguments or browser automation steps.

lgs-search