The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and process external data sources including session corpora, transcripts, and documentation to mine persona traits. Because this external data is processed and used to influence subsequent agent actions (such as synthesis and drafting), it presents a surface for indirect prompt injection if the source materials contain adversarial instructions.
Ingestion points: Step 4a and 4b in SKILL.md identify corpora, transcripts, and documents as primary inputs.
Boundary markers: The instructions do not specify explicit delimiters or "ignore" instructions for the ingested evidence.
Capability inventory: The skill utilizes bash utilities (jq, grep, sed), python3, and specialized agent tools for processing and validation.
Sanitization: No specific sanitization or filtering of the input text is described before it is processed by the analysis agents.
Command Execution: The skill involves running shell commands using bash and utilities like jq, grep, and sed to perform deterministic extraction of text from session files. These are standard data processing techniques but involve direct interaction with the local file system and shell environment.
Dynamic Execution Pattern: Step 7 utilizes a python3 one-liner to parse YAML frontmatter and calculate the length of the description field. This involves the runtime execution of a script to perform metadata validation, which is a routine developer task but represents a dynamic execution pattern.

personafy