The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to interpret "odd-looking code patterns" as "load-bearing scar tissue" rather than errors. This creates an attack surface where an attacker could potentially disguise malicious logic to match the "historical patterns" described in this archaeology, causing the agent to ignore or protect the code from modification.
Ingestion points: Analyzes project source code (src/) and bug symptoms.
Boundary markers: None present.
Capability inventory: Provides diagnostic commands for code analysis and pattern matching.
Sanitization: None.
[COMMAND_EXECUTION]: The documentation provides multiple grep command patterns for diagnostic purposes (e.g., grep -rn "CharacterAutoLoads" *.project.json and grep -rn "DEBUG_" src/Server). These are standard development tools and are used here for instructional purposes.
[EXTERNAL_DOWNLOADS]: Mentions the dig1t/maid package as a dependency. As the skill author is identified as dig1t, this is documented as a legitimate vendor-owned resource.

roblox-failure-archaeology