The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute a series of repository management and automation tasks.
It invokes git for pulling updates (git pull), switching branches (git checkout), and committing changes.
It utilizes the GitHub CLI (gh) for creating Pull Requests (gh pr create) and performing automated merges (gh pr merge).
It uses yq to programmatically extract configuration values from YAML files.
[DYNAMIC_EXECUTION]: The instructions direct the agent to generate and execute shell scripts based on values parsed from local configuration files.
In Phase 0.2, the agent constructs a loop to iterate through repository paths and execute git commands, effectively assembling executable shell code from configuration data at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection because it incorporates external, untrusted content into its workflow.
Ingestion points: The skill reads research data from external sources via a delegated worker (harness:delegate) and processes local context files.
Boundary markers: The delegation requests include constraints blocks intended to guide sub-worker behavior, but these do not provide robust isolation from malicious content in the ingested data.
Capability inventory: The skill possesses capabilities that could be abused if the agent is influenced by malicious input, including file writing, committing code to a repository, and merging Pull Requests.
Sanitization: There is no evidence of explicit sanitization or content filtering for the research findings before they are included in the synthesized report.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch data from remote services.
It uses git pull to download updates from configured repositories.
It interacts with GitHub APIs through the gh CLI for Pull Request management. These operations target well-known and trusted services.

daily-research