The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests various untrusted external project files, which creates an attack surface where malicious instructions could be embedded to influence the agent's estimation or behavior.\n
Ingestion points: The process involves reading design.md, tasks.md, AGENTS.md, delta.md, project wiki pages, and the existing repository source code and tests.\n
Boundary markers: There are no instructions in the skill to use delimiters (such as XML tags or triple quotes) or explicit warnings to ignore embedded instructions when interpolating file content into the prompt.\n
Capability inventory: The skill calls read-codebase-context, performs local file reads and writes (limited to documentation directories and INDEX.md), and executes git diff --check.\n
Sanitization: No mention of content sanitization, filtering, or validation is provided for the ingested data.

estimate-feature