The Agent Skills Directory

[PROMPT_INJECTION]: The skill analyzes untrusted text files (prose, documentation) without implementing boundary markers to isolate file content from the agent's instruction stream.
Ingestion points: Target files are read and processed using the Read and Grep tools during the detection workflow (SKILL.md).
Boundary markers: The skill lacks explicit delimiters or instructions to treat ingested file content as non-executable data.
Capability inventory: The skill utilizes Read, Grep, and TodoWrite tools, and can trigger external skill execution through scribe:doc-generator (SKILL.md).
Sanitization: No sanitization or filtering logic is present to prevent the agent from being influenced by instructions embedded within the scanned files.
[DATA_EXFILTRATION]: The configuration module in modules/config-file.md supports an extends field that allows the agent to load and merge configuration data from arbitrary file paths. This could be exploited to trick the agent into reading sensitive files outside of the repository's context.
[COMMAND_EXECUTION]: The skill provides logic for the agent to execute shell-based commands like grep, awk, and wc to perform structural analysis on target files (modules/structural-patterns.md). This allows the agent to interact with the local file system based on paths encountered during the scanning process.

slop-detector