The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install the sumble-pp-cli tool via npx -y @mvanhorn/printing-press-library and go install github.com/mvanhorn/printing-press-library/.... These methods fetch executable content from the vendor's repositories.
[COMMAND_EXECUTION]: The skill facilitates the execution of the sumble-pp-cli binary using the Read Bash tool, allowing the agent to interact with the system shell.
[DATA_EXFILTRATION]: The CLI supports the --deliver webhook:<url> flag, which sends command results to an arbitrary URL via a POST request. Additionally, the feedback command can be configured to transmit local information to a remote endpoint if the SUMBLE_FEEDBACK_ENDPOINT environment variable is defined.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to processing malicious instructions embedded in external data sources.
Ingestion points: Data enters the context from local CSV files provided to the reconcile command and from responses fetched from the Sumble API (e.g., organizations find, people find).
Boundary markers: No explicit boundary markers or instructions to ignore embedded prompts are present in the skill definition.
Capability inventory: The skill can execute CLI commands, write results to local files using --deliver file:<path>, and perform network requests via --deliver webhook:<url>.
Sanitization: There is no documentation regarding the validation or sanitization of data retrieved from external sources before processing.

pp-sumble