The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources including issue tracker bodies and Git commit logs, creating a potential vector for indirect prompt injection.
Ingestion points: The skill reads issue content via backlog_view in SKILL.md and reads task criteria via plan read in references/migration-fidelity-gate.md.
Capability inventory: The skill possesses high-impact capabilities including repository modification (git commit, git push), issue resolution (backlog resolve), and the ability to author and dispatch further plans via sam_plan.
Boundary markers: While the skill follows structured Mermaid diagrams for logic, it does not instruct sub-agents to treat interpolated external data as untrusted or provide delimiters to separate instructions from data.
Sanitization: No explicit sanitization or filtering is applied to the {body} or {acceptance_criteria} strings before they are embedded into sub-agent prompts.
[COMMAND_EXECUTION]: The skill performs several command-line operations to manage the development lifecycle.
Git operations: Executes git log, git diff, and git push to manage code changes and history.
Tool integration: Utilizes a custom CLI tool (sam_cli) for plan orchestration, backlog updates, and quality gate enforcement.
Script execution: The skill invokes a vendor-provided Python script (verify_migration_fidelity.py) using uv run as part of its migration sign-off procedure.

complete-implementation