The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructions in Step 2 direct the agent to retrieve and execute specific shell commands (such as a "Locate command" and "References to a symbol command") defined within the repository's tooling.md file. This pattern allows the project's own data to control the agent's executable actions, which could be exploited if the file is maliciously modified.
[INDIRECT_PROMPT_INJECTION]: The skill processes numerous files from the repository to determine its actions, creating a surface for indirect prompt injection.
Ingestion points: The skill reads tooling.md, progress-tracker.md, decision-log.md, build-plan.md, ui-registry.md, library-docs.md, design-registry.md, glossary.md, and the entire repository source code via the Grep and Read tools.
Boundary markers: No boundary markers are mentioned; the instructions do not include delimiters or specific warnings to ignore instructions that might be embedded in the analyzed data.
Capability inventory: The skill uses the Bash tool for git operations, database queries, and executing the commands retrieved from tooling.md. It also uses Write and Edit tools to modify project documentation.
Sanitization: There is no mention of sanitization or validation for the commands extracted from tooling.md or for the data used to construct database queries.
[COMMAND_EXECUTION]: The skill performs multiple shell operations, including:
Git operations such as git fetch, git diff, and git merge-base to scope changes.
Database interactions where the agent is instructed to "Query the real database" to verify data-related tasks.
Execution of the commands discovered in the tooling.md file.

dev-sync