The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands to analyze code changes and search for interaction history.
Evidence: references/shared/change-analysis.md instructs the agent to run git status, git diff, and git log to scope changes.
Evidence: references/shared/prompt-recovery.md utilizes grep, head, sqlite3, and gh to find and peak into history files.
[DATA_EXFILTRATION]: The skill is designed to locate and read sensitive agent interaction history from numerous application directories.
Evidence: references/shared/prompt-recovery.md searches through directories such as ~/.claude/projects, ~/.codex/sessions, ~/.continue/sessions, ~/.gemini/tmp, and VS Code global storage paths.
Evidence: It attempts to query local SQLite databases (state.vscdb) to extract chat and composer history.
[PROMPT_INJECTION]: The skill processes untrusted data from code repositories and agent logs which may contain malicious instructions.
Ingestion points: Untracked files discovered via git status, scoped git diffs, and historical session logs in JSON, JSONL, or SQLite formats.
Boundary markers: There are no explicit instructions to use delimiters or ignore embedded instructions when reading external content.
Capability inventory: The skill has the ability to execute shell commands (git, grep, sqlite3) and write generated articles to the filesystem (docs/articles/).
Sanitization: No sanitization or validation logic is defined to prevent the execution of instructions embedded in the analyzed code or history.

article