The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on several local Python scripts to handle its internal logic. These include scripts/resolve_spec_paths.py for directory handling, scripts/merge_state.py for managing workflow state, and scripts/count_tasks.py for progress tracking. These scripts are invoked as part of the skill's primary routing and core rules.
[INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface where it ingests untrusted content from the project being managed, which could potentially influence its coordination logic.
Ingestion points: The skill is instructed to read settings from .claude/ralph-specum.local.md and process various spec artifacts (research, requirements, design) from the local directory.
Boundary markers: There are no explicit delimiters or boundary markers defined in the instructions for isolating content read from project-local configuration files.
Capability inventory: The skill has capabilities to execute local scripts, perform file system operations (writing .progress.md and .ralph-state.json), and manipulate the repository structure via branch and worktree creation.
Sanitization: The skill instructions do not specify any validation or sanitization routines for the data ingested from the .local.md settings file.

ralph-specum