The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's workflow (SKILL.md, Step 3) explicitly instructs the agent to execute five local validation scripts: scripts/test_maven4_model_contract.py, scripts/test_dependency_property_layout.py, scripts/test_bom_import_conflicts.py, scripts/test_dependency_alignment.py, and scripts/check-bom-alignment.sh. These scripts are mentioned as 'Validation Gates' but their source code is not provided in the skill manifest, making their execution unverifiable and potentially risky if an agent attempts to locate or create them.
[COMMAND_EXECUTION]: The skill requires the execution of shell-based Maven commands, including mvn help:effective-pom, mvn dependency:go-offline, and mvn compile. These commands can trigger the execution of arbitrary Maven plugins and build logic defined in external project files.
[DYNAMIC_EXECUTION]: The skill relies on the runtime execution of Python and Shell scripts to validate Maven model contracts and dependency alignments. This behavior is central to the skill's validation process.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection due to its core function of processing untrusted build data.
Ingestion points: The agent is instructed to read and analyze pom.xml files, Maven metadata, and generated effective POMs from the project environment (Workflow Step 4).
Boundary markers: There are no explicit instructions or delimiters defined to separate untrusted project data from the agent's internal instructions or prompts.
Capability inventory: The skill has the capability to execute shell commands (mvn) and run local scripts (python, sh) based on the findings of its analysis.
Sanitization: No sanitization, escaping, or validation logic is defined for the external project content before it is processed or used in shell commands.

ddd4j-bom