The Agent Skills Directory

[SAFE]: The skill functions as a documentation maintenance tool, specifically scoped to reconcile information across durable context files such as PRODUCT.md, GLOSSARY.md, and docs/decisions/. It lacks high-risk capabilities like network access, arbitrary command execution, or credential handling.
[SAFE]: The instructions include robust safety constraints, explicitly forbidding the agent from acting as a decision-maker or changing product code. It mandates that the agent leave sources unchanged when intent is unclear and requires asking explicit questions to resolve conflicts.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from project-level documentation (PRODUCT.md, GLOSSARY.md, etc.), which represents a potential attack surface for indirect prompt injection. However, this is the core intended purpose of the skill, and the behavioral guardrails (e.g., "Preserve authority", "never choose by file date... or silence") effectively mitigate the risk of the agent being misled by instructions embedded within those documents. The skill's actions are limited to updating documentation, further minimizing potential impact.
[SAFE]: No obfuscation, persistence mechanisms, or privilege escalation patterns were detected. The skill uses standard markdown and JSON for its instructions and evaluation fixtures, showing no evidence of malicious intent or hidden behaviors.

maintain-project-context