The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes Architecture Decision Records (ADRs) and Product Decision Records (PDRs) which are stored as markdown files within the repository. This represents a vulnerability surface where instructions embedded in these records could potentially influence the agent's behavior during consistency checks or Slack reporting.
Ingestion points: Reads content from docs/adr/*.md and docs/pdr/*.md to check for contradictions and duplications.
Boundary markers: Absent. The skill does not explicitly instruct the agent to ignore or delimit potentially untrusted content within the records themselves.
Capability inventory: The skill has the capability to write files (creating/updating records and the decision queue) and execute network requests via curl to the Slack API.
Sanitization: Absent. Content from the records is interpolated into Slack message templates and file updates without visible escaping or validation mechanisms.
[COMMAND_EXECUTION]: The documentation provides explicit shell command templates using curl to interact with the Slack API. While these are presented as implementation examples for the agent's integration, they demonstrate a capability to execute arbitrary network-oriented shell commands.
[DATA_EXFILTRATION]: The skill is designed to transmit project decision metadata, author names, and status updates to external Slack channels via https://slack.com/api/chat.postMessage. While Slack is a well-known service and this behavior is core to the skill's purpose, it involves the transmission of internal organizational data to an external API endpoint using a provided SLACK_BOT_TOKEN.

decision-record-governance