[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6996QG8NWMDC83GTK885TDD":3,"skill-same:01M6996QG8NWMDC83GTK885TDD":61},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":32,"linkOnly":33,"markdownContent":34,"markdownBody":35,"frontMatter":36,"refName":37,"commitSha":38,"skillDirectory":39,"forksCount":40,"repositoryHtmlUrl":41,"repositoryHomepageUrl":42,"ownerHtmlUrl":43,"ownerBlogUrl":44,"ownerDescription":45,"updatedAt":46,"install":47,"resources":60},"01M6996QG8NWMDC83GTK885TDD","fuzzing-python-expert",null,"Use this skill to fuzz open source Python software projects using Atheris.","google\u002Foss-fuzz","google","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F1342004?v=4","01M698VYEFDM0VJBZDYYJ2T83S","Google","01M6996E0MR95P8E5T4FN1KBHN",12952,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Foss-fuzz\u002Fblob\u002Fmaster\u002Finfra\u002Fexperimental\u002Fagent-skills\u002Ffuzzing-python-expert\u002FSKILL.md",[],[19,20,21,22,23,24,25,26,27,28,29,30,31],"python","fuzzing","fuzzing python expert","python expert","expert","fuzzing python","open source","open","projects","use","software","atheris","source","2026-11-10 10:12:30",false,"---\nname: fuzzing-python-expert\ndescription:\n  Use this skill to fuzz open source Python software projects using Atheris.\n---\n\n# Fuzzing Python expert\n\nThis skill provides the agent with the knowledge and tools to write, build, and\nvalidate fuzz targets for Python projects integrated into OSS-Fuzz. Python\nfuzzing uses [Atheris](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fatheris), which wraps libFuzzer\nand instruments Python bytecode for coverage-guided fuzzing.\n\n## Fundamental Concepts\n\n### OSS-Fuzz base image\n\nPython projects must use the Python base builder image:\n\n```dockerfile\nFROM gcr.io\u002Foss-fuzz-base\u002Fbase-builder-python\n```\n\nSet `language: python` in `project.yaml`.\n\n### Harness structure\n\nA Python fuzz target is a `.py` file (named `fuzz_\u003Ctarget>.py` by convention)\nthat follows this pattern:\n\n```python\n#!\u002Fusr\u002Fbin\u002Fpython3\nimport sys\nimport atheris\n\n# Import the module under test after atheris.instrument_imports() or within\n# the atheris.instrument_all() context so bytecode is instrumented.\n\ndef TestOneInput(data):\n    fdp = atheris.FuzzedDataProvider(data)\n    # Extract typed values from the raw fuzzer bytes.\n    value = fdp.ConsumeString(128)\n    try:\n        my_module.parse(value)\n    except (ValueError, TypeError, KeyError):\n        # Expected exceptions from invalid input are not bugs.\n        pass\n\ndef main():\n    atheris.instrument_all()   # instrument all loaded Python modules\n    atheris.Setup(sys.argv, TestOneInput, enable_python_coverage=True)\n    atheris.Fuzz()\n\nif __name__ == \"__main__\":\n    main()\n```\n\n### Instrumenting imports selectively\n\nWhen `instrument_all()` is too broad (e.g. causes conflicts with C extensions),\ninstrument specific modules using the `instrument_imports()` context manager:\n\n```python\nimport atheris\nwith atheris.instrument_imports():\n    import my_module\n    import my_module.subpackage\n```\n\n### FuzzedDataProvider reference\n\n`atheris.FuzzedDataProvider` splits the raw byte stream into typed values:\n\n| Method | Description |\n|---|---|\n| `ConsumeBytes(count)` | `bytes` of length count |\n| `ConsumeByteList(count)` | `list[int]` of length count |\n| `ConsumeString(count)` | decoded `str` (may contain surrogates) |\n| `ConsumeUnicode(count)` | `str` without surrogates |\n| `ConsumeUnicodeNoSurrogates(count)` | strict `str` |\n| `ConsumeInt(nbytes)` | signed int from nbytes |\n| `ConsumeIntInRange(min, max)` | int in range |\n| `ConsumeFloat()` | float |\n| `ConsumeBool()` | bool |\n| `ConsumeIntList(count, nbytes)` | list of ints |\n| `PickValueInList(lst)` | random element |\n| `ConsumeRemainingBytes()` | all remaining bytes |\n\n### Building in OSS-Fuzz\n\n`build.sh` installs the target package and uses the `compile_python_fuzzer`\nhelper to turn each `fuzz_*.py` file into a standalone fuzzer binary in `$OUT`:\n\n```bash\n# build.sh\n\n# Install the package under test.\npip3 install .\n\n# Compile all fuzz targets found in $SRC.\nfor fuzzer in $(find $SRC -name 'fuzz_*.py'); do\n    compile_python_fuzzer \"$fuzzer\"\ndone\n```\n\n`compile_python_fuzzer` handles linking against Atheris and libFuzzer and\nproduces an executable in `$OUT` named after the `.py` file.\n\n### Seed corpus and dictionaries\n\n- Place seed files in `$OUT\u002F\u003Cfuzzer_name>_seed_corpus\u002F` or zip them as\n  `$OUT\u002F\u003Cfuzzer_name>_seed_corpus.zip`.\n- Dictionaries go to `$OUT\u002F\u003Cfuzzer_name>.dict` — especially valuable for\n  text-format parsers (JSON, XML, YAML, CSV, etc.).\n- For targets that parse a structured format, generating seeds with a script\n  beats hand-picking a few files — random mutation rarely passes the parser's\n  early checks. See the [structured seed generation\n  reference](..\u002Foss-fuzz-engineer\u002Freferences\u002Fstructured_seed_generation.md).\n\n## Characteristics of good Python fuzzing harnesses\n\n1. **Instruments the right modules**: use `instrument_all()` for simplicity or\n   `instrument_imports()` for targeted instrumentation. Without instrumentation\n   coverage guidance is blind.\n2. **Targets attack surface**: parsers, template engines, serialisation\n   (pickle, JSON, YAML, XML), network protocol handling, and any API that\n   accepts untrusted strings or bytes.\n3. **Catches expected exceptions**: wrap target calls in `try\u002Fexcept` for the\n   documented exception types the target raises on bad input. Only unexpected\n   exceptions and hard crashes are findings.\n4. **Uses `FuzzedDataProvider` for structured input** rather than feeding raw\n   bytes directly to APIs that expect text — most Python APIs work on strings,\n   not bytes.\n5. **Avoids importing under test inside `TestOneInput`**: imports should happen\n   at module level (inside the `instrument_imports()` block if used) so they\n   are instrumented and not re-executed per iteration.\n6. **Is fast**: avoid file I\u002FO, network calls, or subprocess invocations inside\n   `TestOneInput`.\n7. **Avoids non-determinism**: no `random`, no `datetime.now()`, no\n   `os.urandom()` inside the fuzz function.\n8. **Enables Python coverage**: pass `enable_python_coverage=True` to\n   `atheris.Setup` for bytecode-level coverage tracking.\n\n## What Python fuzzing finds\n\nPython is memory-safe, so the focus is on:\n\n- **Unhandled exceptions**: `ValueError`, `RecursionError`, `MemoryError`,\n  `UnicodeDecodeError`, and any exception the library should have caught and\n  converted to a clean error.\n- **`AssertionError`**: internal invariant violations triggered by crafted\n  input.\n- **Hangs \u002F infinite loops**: detected by OSS-Fuzz's timeout.\n- **Crashes in C extensions**: Python libraries that wrap C code (e.g. lxml,\n  Pillow, cryptography) can still have memory-corruption bugs in their C layer,\n  which Atheris will surface because libFuzzer runs the whole process.\n- **Logic errors**: incorrect output, silent data corruption, wrong parsing\n  results on edge-case inputs.\n\n## Operational guidelines\n\n- Always validate with:\n  ```\n  python3 infra\u002Fhelper.py build_fuzzers \u003Cproject>\n  python3 infra\u002Fhelper.py check_build \u003Cproject>\n  python3 infra\u002Fhelper.py run_fuzzer \u003Cproject> \u003Cfuzzer_name> -- -max_total_time=30\n  ```\n- An instant crash usually means an exception is being raised on every input.\n  Run the harness locally (`python3 fuzz_target.py`) on a sample input to\n  debug before building through OSS-Fuzz.\n- Test the harness locally before the OSS-Fuzz build:\n  ```python\n  python3 fuzz_target.py \u003Cseed_file>\n  ```\n  Atheris supports running in single-input mode outside libFuzzer.\n- Install the package locally (`pip3 install .`) and iterate quickly before\n  going through the Docker build.\n- When iterating locally clone the upstream repo and switch the Dockerfile from\n  `RUN git clone` to `COPY` to avoid network round-trips.\n- Document why each entry point was chosen and what class of bugs it may find.\n","# Fuzzing Python expert\n\nThis skill provides the agent with the knowledge and tools to write, build, and\nvalidate fuzz targets for Python projects integrated into OSS-Fuzz. Python\nfuzzing uses [Atheris](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fatheris), which wraps libFuzzer\nand instruments Python bytecode for coverage-guided fuzzing.\n\n## Fundamental Concepts\n\n### OSS-Fuzz base image\n\nPython projects must use the Python base builder image:\n\n```dockerfile\nFROM gcr.io\u002Foss-fuzz-base\u002Fbase-builder-python\n```\n\nSet `language: python` in `project.yaml`.\n\n### Harness structure\n\nA Python fuzz target is a `.py` file (named `fuzz_\u003Ctarget>.py` by convention)\nthat follows this pattern:\n\n```python\n#!\u002Fusr\u002Fbin\u002Fpython3\nimport sys\nimport atheris\n\n# Import the module under test after atheris.instrument_imports() or within\n# the atheris.instrument_all() context so bytecode is instrumented.\n\ndef TestOneInput(data):\n    fdp = atheris.FuzzedDataProvider(data)\n    # Extract typed values from the raw fuzzer bytes.\n    value = fdp.ConsumeString(128)\n    try:\n        my_module.parse(value)\n    except (ValueError, TypeError, KeyError):\n        # Expected exceptions from invalid input are not bugs.\n        pass\n\ndef main():\n    atheris.instrument_all()   # instrument all loaded Python modules\n    atheris.Setup(sys.argv, TestOneInput, enable_python_coverage=True)\n    atheris.Fuzz()\n\nif __name__ == \"__main__\":\n    main()\n```\n\n### Instrumenting imports selectively\n\nWhen `instrument_all()` is too broad (e.g. causes conflicts with C extensions),\ninstrument specific modules using the `instrument_imports()` context manager:\n\n```python\nimport atheris\nwith atheris.instrument_imports():\n    import my_module\n    import my_module.subpackage\n```\n\n### FuzzedDataProvider reference\n\n`atheris.FuzzedDataProvider` splits the raw byte stream into typed values:\n\n| Method | Description |\n|---|---|\n| `ConsumeBytes(count)` | `bytes` of length count |\n| `ConsumeByteList(count)` | `list[int]` of length count |\n| `ConsumeString(count)` | decoded `str` (may contain surrogates) |\n| `ConsumeUnicode(count)` | `str` without surrogates |\n| `ConsumeUnicodeNoSurrogates(count)` | strict `str` |\n| `ConsumeInt(nbytes)` | signed int from nbytes |\n| `ConsumeIntInRange(min, max)` | int in range |\n| `ConsumeFloat()` | float |\n| `ConsumeBool()` | bool |\n| `ConsumeIntList(count, nbytes)` | list of ints |\n| `PickValueInList(lst)` | random element |\n| `ConsumeRemainingBytes()` | all remaining bytes |\n\n### Building in OSS-Fuzz\n\n`build.sh` installs the target package and uses the `compile_python_fuzzer`\nhelper to turn each `fuzz_*.py` file into a standalone fuzzer binary in `$OUT`:\n\n```bash\n# build.sh\n\n# Install the package under test.\npip3 install .\n\n# Compile all fuzz targets found in $SRC.\nfor fuzzer in $(find $SRC -name 'fuzz_*.py'); do\n    compile_python_fuzzer \"$fuzzer\"\ndone\n```\n\n`compile_python_fuzzer` handles linking against Atheris and libFuzzer and\nproduces an executable in `$OUT` named after the `.py` file.\n\n### Seed corpus and dictionaries\n\n- Place seed files in `$OUT\u002F\u003Cfuzzer_name>_seed_corpus\u002F` or zip them as\n  `$OUT\u002F\u003Cfuzzer_name>_seed_corpus.zip`.\n- Dictionaries go to `$OUT\u002F\u003Cfuzzer_name>.dict` — especially valuable for\n  text-format parsers (JSON, XML, YAML, CSV, etc.).\n- For targets that parse a structured format, generating seeds with a script\n  beats hand-picking a few files — random mutation rarely passes the parser's\n  early checks. See the [structured seed generation\n  reference](..\u002Foss-fuzz-engineer\u002Freferences\u002Fstructured_seed_generation.md).\n\n## Characteristics of good Python fuzzing harnesses\n\n1. **Instruments the right modules**: use `instrument_all()` for simplicity or\n   `instrument_imports()` for targeted instrumentation. Without instrumentation\n   coverage guidance is blind.\n2. **Targets attack surface**: parsers, template engines, serialisation\n   (pickle, JSON, YAML, XML), network protocol handling, and any API that\n   accepts untrusted strings or bytes.\n3. **Catches expected exceptions**: wrap target calls in `try\u002Fexcept` for the\n   documented exception types the target raises on bad input. Only unexpected\n   exceptions and hard crashes are findings.\n4. **Uses `FuzzedDataProvider` for structured input** rather than feeding raw\n   bytes directly to APIs that expect text — most Python APIs work on strings,\n   not bytes.\n5. **Avoids importing under test inside `TestOneInput`**: imports should happen\n   at module level (inside the `instrument_imports()` block if used) so they\n   are instrumented and not re-executed per iteration.\n6. **Is fast**: avoid file I\u002FO, network calls, or subprocess invocations inside\n   `TestOneInput`.\n7. **Avoids non-determinism**: no `random`, no `datetime.now()`, no\n   `os.urandom()` inside the fuzz function.\n8. **Enables Python coverage**: pass `enable_python_coverage=True` to\n   `atheris.Setup` for bytecode-level coverage tracking.\n\n## What Python fuzzing finds\n\nPython is memory-safe, so the focus is on:\n\n- **Unhandled exceptions**: `ValueError`, `RecursionError`, `MemoryError`,\n  `UnicodeDecodeError`, and any exception the library should have caught and\n  converted to a clean error.\n- **`AssertionError`**: internal invariant violations triggered by crafted\n  input.\n- **Hangs \u002F infinite loops**: detected by OSS-Fuzz's timeout.\n- **Crashes in C extensions**: Python libraries that wrap C code (e.g. lxml,\n  Pillow, cryptography) can still have memory-corruption bugs in their C layer,\n  which Atheris will surface because libFuzzer runs the whole process.\n- **Logic errors**: incorrect output, silent data corruption, wrong parsing\n  results on edge-case inputs.\n\n## Operational guidelines\n\n- Always validate with:\n  ```\n  python3 infra\u002Fhelper.py build_fuzzers \u003Cproject>\n  python3 infra\u002Fhelper.py check_build \u003Cproject>\n  python3 infra\u002Fhelper.py run_fuzzer \u003Cproject> \u003Cfuzzer_name> -- -max_total_time=30\n  ```\n- An instant crash usually means an exception is being raised on every input.\n  Run the harness locally (`python3 fuzz_target.py`) on a sample input to\n  debug before building through OSS-Fuzz.\n- Test the harness locally before the OSS-Fuzz build:\n  ```python\n  python3 fuzz_target.py \u003Cseed_file>\n  ```\n  Atheris supports running in single-input mode outside libFuzzer.\n- Install the package locally (`pip3 install .`) and iterate quickly before\n  going through the Docker build.\n- When iterating locally clone the upstream repo and switch the Dockerfile from\n  `RUN git clone` to `COPY` to avoid network round-trips.\n- Document why each entry point was chosen and what class of bugs it may find.\n",{"name":5,"description":7},"master","e83ae3515d386e253ed7d86395d8cbdebd3e0917","infra\u002Fexperimental\u002Fagent-skills\u002Ffuzzing-python-expert",2997,"https:\u002F\u002Fgithub.com\u002Fgoogle\u002Foss-fuzz","https:\u002F\u002Fgoogle.github.io\u002Foss-fuzz","https:\u002F\u002Fgithub.com\u002Fgoogle","https:\u002F\u002Fopensource.google\u002F","Google ❤️ Open Source","2026-11-14 23:27:56",{"source":8,"skillName":5,"commands":48,"prompt":58,"zipAvailable":59},[49,52,55],{"tool":50,"command":51},"npx","npx skills add google\u002Foss-fuzz --skill fuzzing-python-expert",{"tool":53,"command":54},"bunx","bunx skills add google\u002Foss-fuzz --skill fuzzing-python-expert",{"tool":56,"command":57},"pnpm","pnpm dlx skills add google\u002Foss-fuzz --skill fuzzing-python-expert","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fgoogle\u002Foss-fuzz\u002Fblob\u002Fmaster\u002Finfra\u002Fexperimental\u002Fagent-skills\u002Ffuzzing-python-expert\u002FSKILL.md",true,[],[62,75,85,96,108],{"skillNo":63,"name":64,"nameZh":6,"description":65,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":32},"01M6996V9MBNRZ17CW93P54CQP","oss-fuzz-engineer","Use this skill to interact with the OSS-Fuzz infrastructure.","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Foss-fuzz\u002Fblob\u002Fmaster\u002Finfra\u002Fexperimental\u002Fagent-skills\u002Foss-fuzz-engineer\u002FSKILL.md",[],[69,70,71,72,73,74,28],"oss fuzz","oss fuzz engineer","engineer","oss","with","infrastructure",{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":32},"01M6996SBDQB83B9ACW5SPSM78","fuzzing-rust-expert","Use this skill to fuzz open source Rust software projects.","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Foss-fuzz\u002Fblob\u002Fmaster\u002Finfra\u002Fexperimental\u002Fagent-skills\u002Ffuzzing-rust-expert\u002FSKILL.md",[],[82,20,23,83,84,25,26,27,28,29,31],"fuzzing rust expert","rust","rust expert",{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":32},"01M6996NKJDXKEJ3TZJZNKXV8M","fuzzing-memory-unsafe-expert","Use this skill to fuzz open source C\u002FC++ software projects.","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Foss-fuzz\u002Fblob\u002Fmaster\u002Finfra\u002Fexperimental\u002Fagent-skills\u002Ffuzzing-memory-unsafe-expert\u002FSKILL.md",[],[92,20,93,23,94,31,29,25,26,27,28,95],"fuzzing memory unsafe expert","memory","unsafe","c++",{"skillNo":97,"name":98,"nameZh":6,"description":99,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":100,"categories":101,"keywords":102,"publishedAt":32},"01M6996KFJ5JV7GTDHH4J4ZTJC","fuzzing-jvm-expert","Use this skill to fuzz open source JVM projects (Java, Kotlin, Scala, etc.) using Jazzer.","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Foss-fuzz\u002Fblob\u002Fmaster\u002Finfra\u002Fexperimental\u002Fagent-skills\u002Ffuzzing-jvm-expert\u002FSKILL.md",[],[103,23,104,20,26,105,106,25,27,107,28,31],"jvm","fuzzing jvm expert","scala","java","kotlin",{"skillNo":109,"name":110,"nameZh":6,"description":111,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":112,"categories":113,"keywords":114,"publishedAt":32},"01M6996H8NDKT37EGSJ31W9YRT","fuzzing-go-expert","Use this skill to fuzz open source Go software projects.","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Foss-fuzz\u002Fblob\u002Fmaster\u002Finfra\u002Fexperimental\u002Fagent-skills\u002Ffuzzing-go-expert\u002FSKILL.md",[],[115,23,116,117,20,25,26,27,28,29,31],"go expert","go","fuzzing go expert"]