The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires installing the gdb-cli package from PyPI or directly from a third-party GitHub repository (https://github.com/Cerdore/gdb-cli.git). The source repository and package are not from a verified or trusted organization.
[INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting and analyzing external, potentially untrusted data such as application binaries and core dumps.
Ingestion points: Data is loaded into the agent's context through the load --binary and attach --pid commands described in SKILL.md.
Boundary markers: The instructions include safety notes advising against GDB auto-loading and the evaluation of functions on untrusted data, which acts as a manual boundary but does not programmatically restrict the agent.
Capability inventory: The skill provides commands like eval-cmd, memory, and thread-apply which allow the agent to inspect and execute logic based on the contents of the loaded data.
Sanitization: There is no evidence of automated sanitization or filtering of the binary or memory content before it is processed by the agent.

gdb-cli