The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes external data from Convex deployments, such as performance insights and log entries, which are then used to guide code analysis and automated fix generation. This creates a surface where adversarially crafted data within the deployment could influence the agent's findings.
Ingestion points: Runtime data is fetched from the deployment via MCP tools including insights, logs, and tables (SKILL.md).
Boundary markers: The instructions do not define explicit delimiters or instructions to ignore potential commands embedded within the logs or performance data.
Capability inventory: The skill has the capability to read project source code, emit findings to a 'findings bus', and suggest/apply code fixes (SKILL.md).
Sanitization: There is no mention of sanitizing or validating the ingested log or insight content before it is processed for root-cause analysis.
[EXTERNAL_DOWNLOADS]: The skill references the @convex-dev/sharded-counter package as a recommended fix for write contention. This is a reference to a resource from a well-known service provider in the Convex ecosystem.
[COMMAND_EXECUTION]: The workflow involves running a deploy-guard step to identify and announce the deployment environment, which typically involves tool-mediated interaction with the deployment CLI or MCP.

convex-advisor