The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses natural instructional language and establishes process gates like 'Native Plan Mode' and 'design approval'. No patterns were found that attempt to override system safety guidelines or extract underlying system prompts.
[DATA_EXPOSURE]: No hardcoded credentials, API keys, or sensitive file paths (e.g., SSH keys, AWS configs) were detected. The skill specifies saving documentation to a docs/ folder within the project, which is standard development behavior.
[OBFUSCATION]: The content is presented in clear, plain text Markdown. No Base64, zero-width characters, homoglyphs, or encoded strings were found. The use of Russian Cyrillic terms (e.g., 'дорожная карта') is consistent with the provided bilingual documentation purpose.
[REMOTE_CODE_EXECUTION]: No remote script downloads or piped shell executions (e.g., curl | bash) were identified. The skill interacts with the user and project files for documentation purposes only.
[PRIVILEGE_ESCALATION]: The instructions do not involve commands like sudo, chmod 777, or modifications to sensitive system directories.
[PERSISTENCE]: There are no attempts to modify shell profiles, cron jobs, or startup services. The skill creates durable project documentation in the repository's docs/ folder as intended.
[METADATA_POISONING]: The YAML frontmatter fields (name, description) accurately reflect the skill's functionality and do not contain hidden instructions.
[INDIRECT_PROMPT_INJECTION]: The skill ingests existing roadmap documents from the project's context. However, it lacks dangerous sinks such as network operations or shell command execution on that data, and the interaction is confined to generating documentation and plans.
[TIME_DELAYED_CONDITIONAL]: No logic was found that gates functionality based on date, time, or environment-specific triggers.
[DYNAMIC_EXECUTION]: The skill does not generate or compile code for runtime execution. It focuses on drafting human-readable Markdown documentation and ticket descriptions.
[DYNAMIC_CONTEXT_INJECTION]: No use of the dynamic command execution syntax (!command) was detected in the SKILL.md file.

writing-roadmaps