The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interacts with the Salla platform through a set of Model Context Protocol (MCP) tools (e.g., app_publish, salla_apps, salla_reference). These tools are used for legitimate administrative tasks related to app store listing management.
[DATA_EXFILTRATION]: The skill handles trial_password and trial_username fields intended for app reviewers. It includes specific security instructions to treat these as test-only credentials, redact them from logs, and never store them in plaintext, which aligns with security best practices for credential handling.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from existing app drafts which could theoretically contain malicious instructions if the draft source was compromised. However, the risk is mitigated by the skill's structured approach and mandatory user confirmation requirements.
Ingestion points: Data is ingested via app_publish action=get which retrieves the current draft from the Salla portal.
Boundary markers: No explicit delimiters are specified for the tool output in the instructions.
Capability inventory: The skill uses the app_publish tool to perform POST, PUT, and GET operations on the Salla publication API.
Sanitization: The skill relies on server-side validation (422 responses) and requires the agent to present a review link to the partner for manual verification before the final submission.

salla-publication-consistency