| name | full-llm-export |
| description | Export repository context into a single plain-text snapshot file for downstream LLM/agent ingestion. Use when asked to generate, refresh, or rebuild `llms-full.txt` (or equivalent repo context dumps), especially before handing codebases to external agents that do not have direct repository context. |
Full LLM Export
Overview
Generate llms-full.txt from repository source files with deterministic include/exclude rules and size limits.
Run Workflow
- Run the exporter script from the target repository root.
- Verify that
llms-full.txt exists and has current content.
- Re-run after meaningful repository changes.
Default command
python .agents/skills/full-llm-export/scripts/export_repo_context.py
Custom root/output
python .agents/skills/full-llm-export/scripts/export_repo_context.py \
--root /path/to/repo \
--output llms-full.txt
Script Behavior
- Include file suffixes:
.py, .md, .rst, .toml, .yaml, .yml, .json, .txt, .ini, .cfg, .sh
- Exclude directories:
.git, .hg, .svn, .venv, venv, __pycache__, .mypy_cache, .pytest_cache, .ruff_cache, node_modules, dist, build, .idea, .vscode
- Exclude lock files:
poetry.lock, package-lock.json, pnpm-lock.yaml, yarn.lock
- Skip files larger than
200_000 bytes.
- Write one concatenated export file with per-file separators and relative paths.
Output Contract
- Default output file:
llms-full.txt in the selected root.
- Header includes:
- export title
- root path
- number of included files
- Each file block includes:
FILE: <relative/path>
- raw text content (best-effort decoding: utf-8, utf-8-sig, latin-1)
Notes
- Prefer running from repository root to avoid accidentally exporting the wrong tree.
- Re-run after major code/docs changes so external agents get fresh context.