The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/finalize_batch.py and the test file tests/test_finalize_batch.py modify sys.path at runtime to include directories from sibling skills (ticket-autopilot and llm-wiki). This allows the skill to dynamically load modules from computed paths relative to its own location.
[COMMAND_EXECUTION]: Instructions in SKILL.md direct the agent to execute shell commands via python3 -B to run internal scripts for ticket emission and batch finalization. Additionally, the test suite tests/test_finalize_batch.py uses subprocess.run to execute git commands and verify CLI behavior.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external specification documents (specs) from the filesystem to generate work tickets. These documents represent an untrusted input surface that could potentially contain malicious instructions intended to manipulate the agent's output or logic during the ticket-slicing process.
Ingestion points: Specification files are read from the project's documentation directories.
Capability inventory: The skill uses shell execution to run Python scripts that modify the filesystem and synchronize data.
Boundary markers: The skill does not explicitly define or enforce markers to isolate instructions within the ingested specs.
Sanitization: There is no evidence of sanitization or filtering applied to the content of the specs before they are processed by the agent.

to-tickets