The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it retrieves and processes untrusted content from external issue trackers to generate task plans for autonomous subagents.\n
Ingestion points: In modules/issue-discovery.md, the skill uses gh and glab CLIs to fetch titles, bodies, and comments from issues.\n
Boundary markers: There are no explicit instructions or delimiters defined to help the agent isolate external issue content from its core instructions.\n
Capability inventory: The skill dispatches subagents with the authority to modify source code, execute shell commands, and interact with the git platform.\n
Sanitization: Extracted requirements are used to form task plans in modules/task-planning.md without visible sanitization or filtering of the retrieved content.\n- [COMMAND_EXECUTION]: The skill executes local shell commands for git operations, platform interaction, and process isolation.\n
Evidence: The skill frequently uses git, gh, glab, and tmux throughout its modules for branch management, issue updates, and parallel execution coordination.\n- [DATA_EXFILTRATION]: The skill facilitates sending feedback to a vendor-controlled public repository.\n
Evidence: modules/completion.md contains instructions to post "Tooling Reflection" feedback to the author's public GitHub discussion page at https://github.com/athola/claude-night-market/discussions. While it warns against including repository-specific code, it creates an automated pathway for observations to leave the local environment.

do-issue