The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/configure-keyapi-auth.mjs is designed to modify the user's shell configuration files, including ~/.bashrc, ~/.zshrc, ~/.profile, and Windows PowerShell profiles. This modification is used to persist the KEYAPI_TOKEN environment variable across all future shell sessions, which serves as a persistence mechanism.
[CREDENTIALS_UNSAFE]: The authentication configuration script scripts/configure-keyapi-auth.mjs accepts the KEYAPI_TOKEN through a command-line argument (--token). This practice is insecure as it can lead to the API token being stored in plain text within shell history logs (e.g., .bash_history or .zsh_history).
[DATA_EXFILTRATION]: The skill performs network requests to api.keyapi.ai and docs.keyapi.ai using the fetch API in scripts/keyapi-api.mjs and scripts/search-keyapi-docs.mjs. These requests transmit request parameters and Bearer tokens to the vendor's infrastructure.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from the LinkedIn API (such as user posts, bios, and job descriptions) into the agent's context.
Ingestion points: Data retrieved from LinkedIn via scripts/keyapi-api.mjs and processed in the analytical workflows defined in references/linkedin-user-rules.md and others.
Boundary markers: None identified; retrieved content is interpolated into analysis without delimiters.
Capability inventory: The agent can execute local shell scripts, perform outbound network calls, and write files to the workspace.
Sanitization: No escaping or instruction-filtering of the API response data is implemented before the agent processes it for reporting.

keyapi-linkedin