The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's setup instructions and scripts recommend executing remote code directly by piping a shell script from GitHub to the shell interpreter.
Evidence: README.md, SKILL.md, and install-skill.sh contain commands like curl -fsSL https://raw.githubusercontent.com/eddmann/strava-cli/main/install.sh | sh.
Context: The resources belong to the author's verified GitHub account and represent legitimate tool functionality.
[EXTERNAL_DOWNLOADS]: Installation scripts download a compiled binary and skill definitions from external repositories at runtime.
Evidence: install.sh downloads the strava binary from github.com and install-skill.sh fetches SKILL.md from raw.githubusercontent.com.
[DATA_EXFILTRATION]: The skill manages and transmits sensitive OAuth tokens used to access user fitness data.
Evidence: src/strava_cli/config.py saves credentials to ~/.config/strava-cli/config.toml and src/strava_cli/commands/routes.py transmits the access token to www.strava.com via the Authorization header.
Context: Access to the local configuration file is restricted to the file owner (0o600 permissions), and network operations are limited to official Strava API endpoints.
[COMMAND_EXECUTION]: Setup scripts perform environment detection and file system operations.
Evidence: install.sh uses uname for OS detection and chmod +x to make the downloaded binary executable.
[PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection through the ingestion of external data.
Ingestion points: src/strava_cli/commands/activities.py (get_comments, get_activity) and src/strava_cli/commands/athlete.py retrieve data from the Strava API.
Boundary markers: Absent; data is returned to the agent context without delimiters or instructions to ignore instructions inside data.
Capability inventory: The skill allows the agent to make network requests, write files (e.g., GPX/TCX route exports in src/strava_cli/commands/routes.py), and execute system commands through the Typer framework.
Sanitization: Absent; content from the API is returned to the agent without validation or sanitization.

strava