The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool extensively for environment setup, branch detection, git operations, and running internal CLI tools from the gstack suite (e.g., gstack-config, gstack-repo-mode, gstack-timeline-log).
[DYNAMIC_EXECUTION]: In its preamble, the skill dynamically executes output from local binaries using eval and source with computed paths (e.g., eval "$(~/.claude/skills/gstack/bin/gstack-slug)"). This is used for session management and configuration loading.
[INDIRECT_PROMPT_INJECTION]: The skill has a large attack surface as it ingests untrusted data from git diffs, commit messages, and external Greptile comments. However, it incorporates specific instructions for its subagents to treat potential attack patterns found in test files as data to be analyzed rather than instructions to be executed.
[PERSISTENCE]: The skill maintains state across sessions by writing to files in ~/.gstack/ (e.g., sessions, analytics, markers like .activated). It also offers to modify the project's CLAUDE.md file to add 'Skill routing' rules, which is done after a user confirmation via AskUserQuestion.

review