The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of GitHub issue titles and bodies. It manages this risk through multiple layers of mitigation:
Sanitization: In scripts/update-root-body.sh, a jq filter is used to escape HTML characters (< to <) and comment terminators (--> to -->) in issue titles before they are interpolated into the tracking issue's markdown body. This prevents malicious titles from breaking the skill's management markers.
Boundary Markers: The skill uses explicit HTML comment markers (e.g., ) to delineate blocks of content it manages, and its scripts are designed to preserve all content outside these markers.
Capability Scope: While the skill can write to issues, its logic for reassigning parents is driven by issue numbers/IDs rather than instructions found within the issue bodies.
[COMMAND_EXECUTION]: The skill executes several internal shell scripts (reassign-sub-issue.sh, update-root-body.sh, and reassign-lib.sh). These scripts are part of the skill package and perform structured operations using the gh and jq CLI tools. The scripts use set -euo pipefail and include rigorous argument validation to prevent command injection.
[EXTERNAL_DOWNLOADS]: The skill communicates with the GitHub API via the official gh CLI tool. All network operations are directed at github.com (a well-known service). The skill does not download or execute scripts from arbitrary remote sources.
[SAFE]: The skill includes an extensive suite of regression tests (tests/) using a mock GitHub stub to verify edge cases such as race conditions, API failures, and stability of issue reassignments. This indicates a high level of engineering maturity and focus on operational safety.

update-issue-tree