The Agent Skills Directory

[PROMPT_INJECTION]: No attempts to override agent behavior, bypass safety filters, or extract system prompts were detected. The instructions focus on structured data extraction and internal documentation updates.
[DATA_EXFILTRATION]: No network operations, hardcoded credentials, or access to sensitive local files (such as SSH keys or AWS credentials) were identified. The skill only interacts with project-specific documentation files (DESIGN.md, findings.md).
[REMOTE_CODE_EXECUTION]: The skill does not perform any external downloads, package installations, or execution of remote scripts.
[COMMAND_EXECUTION]: The skill is restricted to standard file operations (Read, Edit, Glob, Grep) and does not involve the execution of arbitrary shell commands or subprocesses.
[OBFUSCATION]: No hidden content, encoded strings (Base64/Hex), or deceptive character usage (homoglyphs/zero-width characters) were found in the skill text or configuration.
[INDIRECT_PROMPT_INJECTION]: The skill processes project-level data (DESIGN.md) and interpolates it into the agent's attention window via findings.md. While this creates a surface for indirect instructions, it is the primary intended function of the 'design bridge' and does not involve executable capabilities or network access that could be leveraged for malicious purposes.
Ingestion points: DESIGN.md (Step 2)
Boundary markers: Absent (uses standard Markdown headers)
Capability inventory: Edit, Read, Glob, Grep tools used for local file manipulation.
Sanitization: The skill performs heuristic-based extraction but does not explicitly sanitize the content of the design file.

design-bridge