The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted repository content, including README files, Markdown documents, commit messages, and issue logs, which could contain malicious instructions or prompt injection payloads.
Ingestion points: Untrusted data enters the agent context during Step 3 (Establish Current Project Reality), Step 8 (Audit Documentation Before Editing), and during the execution of the capability-sweep check in the scripts/updatedocs.ts script.
Boundary markers: The skill contains robust defensive instructions in references/SECURITY.md (Category 3) that explicitly warn the agent to treat repository Markdown as untrusted and to ignore any embedded directives that attempt to override instructions or exfiltrate data.
Capability inventory: The agent has access to powerful tools including bash, write_to_file, and replace_file_content across all scripts.
Sanitization: The skill implements a 'Zero Secret Exposure Protocol' and a sanitizeClientJargon function in the provided script to mask sensitive tokens, credentials, and internal identifiers.
[COMMAND_EXECUTION]: The skill facilitates the execution of local shell commands using the bash tool to run its internal audit script (scripts/updatedocs.ts). This script uses spawnSync to execute git and bun commands for repository analysis and documentation generation.
[DYNAMIC_EXECUTION]: The skill uses a Bun-based script (scripts/updatedocs.ts) to dynamically generate documentation artifacts such as CLIENT_HANDOFF.md, CLIENT_CHANGELOG.md, and CLIENT_WORK_REPORT.html based on the current repository state and Git history. This is performed using surgical block replacements to preserve existing content structure.

updatedocs