The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple local Python scripts (e.g., harness/scripts/wiki-sync.py, fdk/tools/skill-usage.py, harness/scripts/skill-health.py) and shell utilities like grep, find, and comm. These operations are central to its purpose as a linter but rely on the security and integrity of the scripts located within the project's harness/ and fdk/ directories.
[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Category 8) because it ingests and processes untrusted data from the repository's wiki pages and source code to identify links, contradictions, and debt markers.
Ingestion points: Wiki files (llmwiki/wiki/), source code files (.py, .js, etc.), and metrics logs (harness/metrics/).
Boundary markers: Absent. The skill analyzes raw file content without the use of protective delimiters or instructions to ignore embedded commands.
Capability inventory: The skill has the ability to read/write files and execute arbitrary shell commands and Python scripts.
Sanitization: Absent. No validation or sanitization is performed on the content extracted from the analyzed files before it is processed.
[DATA_EXFILTRATION]: The skill accesses internal metrics and log files (events.jsonl, scratch-log.jsonl) to correlate code changes with wiki updates. This access exposes potentially sensitive session metadata and internal logs to the agent's context, although no external network transmission was identified.

lint