The Agent Skills Directory

[DATA_EXPOSURE]: The skill reads local agent history logs from sensitive directories such as ~/.cursor, ~/.claude, and ~/.codex. These logs contain historical chat data, code snippets, and tool outputs. This data access is essential for the skill's stated purpose of extracting and backfilling knowledge from past sessions into a persistent library.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from local logs which could theoretically contain malicious instructions.
Ingestion points: Historical session logs in JSONL format processed by scripts/parse_agent_logs.py.
Boundary markers: SKILL.md and references/write-criteria.md provide clear instructions to the agent to ignore user prompts and focus only on assistant conclusions and technical facts.
Capability inventory: The skill possesses the ability to write to a knowledge base via MCP tools, execute local analysis scripts, and open a generated HTML report in a browser.
Sanitization: The agent is explicitly instructed to redact secrets, personally identifiable information (PII), and sensitive data before writing any notes.
[COMMAND_EXECUTION]: The skill utilizes dynamic context injection (!command syntax) in SKILL.md to automate environment discovery. These commands (such as find, git remote, and date) are standard development utilities used to determine the correct project paths and generate branch metadata for the knowledge sync.

log-to-dosu-knowledge