The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script harness/scripts/query-log.py for telemetry purposes.
Evidence: Step 8 in SKILL.md uses python3 harness/scripts/query-log.py --record --question "<câu hỏi>" --pages "<slug1,slug2>" --tokens <ước tính token đã đọc> --tier <1|2|3>.
Context: The user-supplied question (<câu hỏi>) is interpolated directly into the command-line arguments, which could lead to command injection if the agent does not properly escape the string.
[COMMAND_EXECUTION]: The skill uses ripgrep (rg) and a custom script retrieval-eval.py for content discovery and verification.
Evidence: Step 1 uses rg -c '<term>' wiki/ and Step 0 (Purpose) mentions retrieval-eval.py --check.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests untrusted data and possesses file-write capabilities.
Ingestion points: Processes data from the raw/ directory (Step 4) and various wiki/ pages.
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore instructions embedded within the source materials.
Capability inventory: The skill has the ability to create new wiki pages, modify wiki/index.md, and write to wiki/log.md (Step 6 and Output Report section).
Sanitization: No sanitization or validation steps are described for the content being synthesized from the raw/ sources before it is committed to the wiki.

query