The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources like git logs and Jira tickets, which could contain malicious instructions intended to influence the agent's summarization or subsequent actions.
Ingestion points: collect.sh (git log messages), tickets-jira.sh (Jira summaries), and sessions.sh (local Claude Code transcripts).
Boundary markers: The agent is instructed to distinguish between observed evidence and its own prose, but the shell scripts do not wrap ingested data in specific safety delimiters.
Capability inventory: The skill has the ability to execute shell scripts, read/write local files, and push commits to remote git repositories.
Sanitization: The sessions.sh script filters for user-originated messages and applies length truncation to minimize the impact of long or potentially malicious inputs.
[COMMAND_EXECUTION]: The skill uses the agent's shell capabilities to execute local bash scripts provided in the skill package for data collection and system keyring access.
[DATA_EXFILTRATION]: The skill accesses sensitive development data, including git history and transcripts. This data is transmitted to a user-configured Jira API endpoint or pushed to a user-defined remote git repository as part of its primary functionality.

standup