The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the olcli command-line tool from vendor-controlled repositories.
Evidence: Installation instructions and scripts/install.sh reference brew tap aloth/tap and the @aloth/olcli npm package.
[COMMAND_EXECUTION]: The skill executes system commands through the olcli utility to perform LaTeX project management tasks such as pulling, pushing, and compiling documents.
Evidence: Instructions in SKILL.md and README.md detail the use of olcli pull, olcli sync, olcli pdf, and olcli output commands.
[CREDENTIALS_UNSAFE]: The skill manages authentication using Overleaf session cookies. The olcli auth command accepts a user-provided cookie which is stored locally for session persistence.
Evidence: SKILL.md and references/API.md describe copying the overleaf_session2 cookie and storing it in ~/.config/olcli-nodejs/config.json or a local .olauth file.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external project data which could theoretically contain malicious instructions; however, the tool is a standard utility wrapper for LaTeX workflows.
Ingestion points: Files downloaded from Overleaf via olcli pull (SKILL.md).
Boundary markers: None identified.
Capability inventory: File system access, network communication via olcli, and archival capabilities via zip (SKILL.md).
Sanitization: Not explicitly documented in the skill instructions.

overleaf