The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled Python script (scripts/check.py) to perform mechanical verification of proposal drafts. This script runs locally within the workspace, focusing on structural errors, citation consistency, and style violations. The execution uses a relative path based on the skill's installation directory and targets filenames generated according to strict internal rules, preventing command injection vectors.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from user-provided seed files and workspace-level configuration files (guidelines.md). While this creates a potential attack surface for indirect prompt injection, the risk is mitigated by the skill's highly specialized and structured instructions. Furthermore, the use of a deterministic verification script ensures that the final output adheres to predefined academic standards regardless of instructions potentially hidden in the input data.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill's operations are confined to the local workspace and the skill's own directory. There are no network operations, hardcoded credentials, or attempts to access sensitive system files (such as SSH keys or environment secrets). The behavior matches the stated purpose of local document generation and revision.

proposal-write