Code
Hub
Workspaces
Following
Trending
Connect
MCP
copy
Create free account
hub
/
github.com/NVIDIA/SkillEvaluator
/ functions
Functions
24,741 in github.com/NVIDIA/SkillEvaluator
⨍
Functions
24,741
◇
Types & classes
1,292
↳
Endpoints
703
↓ 2 callers
Function
_shell_tool
The Claude Code shell tool a shell-tool name stands for: ``PowerShell`` for any spelling of it, else ``Bash``.
src/skillevaluator/plugin_component_risk.py:619
↓ 2 callers
Function
_shift_line_numbers
(line_numbers: frozenset[int], offset: int)
src/skillevaluator/validators/security.py:315
↓ 2 callers
Function
_shipped_paths
Files and symlinks a skill ships, without descending into VCS or cache folders.
src/skillevaluator/harness/overlay.py:263
↓ 2 callers
Function
_show_tier3_progress
()
src/skillevaluator/cli.py:2908
↓ 2 callers
Function
_shown_name
A server name (or other key) for messages: credentials redacted, at most 64 characters.
src/skillevaluator/validators/mcp_static.py:384
↓ 2 callers
Method
_sidecar_compose_model_environment
Retain only non-sensitive task values structurally required by Compose.
src/skillevaluator/tier3/harbor/secure_docker_environment.py:2287
↓ 2 callers
Function
_sidecar_environment_carriers
Map target env names through unpredictable client-safe carriers.
src/skillevaluator/tier3/harbor/secure_docker_environment.py:582
↓ 2 callers
Function
_signals_raise_keyboard_interrupt
While Gym runs, turn SIGTERM/SIGHUP into one KeyboardInterrupt so Gym's cleanup still runs.
src/skillevaluator/tier3/nemo_gym/runner.py:456
↓ 2 callers
Function
_sigterm_handling_harbor_script
Stand-in Harbor CLI that reports readiness and handles SIGTERM itself.
tests/test_harbor_runner_status.py:478
↓ 2 callers
Function
_similarity_result
()
tests/reporting/browser_navigation_fixtures.py:64
↓ 2 callers
Function
_simulate_windows_crt_fstat_identity
Return valid descriptor metadata with a Windows-incompatible identity.
tests/test_harbor_output_provenance.py:62
↓ 2 callers
Function
_single_score_label
(score: float)
src/skillevaluator/reporting/benchmark.py:1786
↓ 2 callers
Function
_single_secret_scan_seconds
Return how long this interpreter takes to redact ``text`` with one secret. Coverage tracing slows the pure-Python scan several times over. Nested
tests/test_harbor_secure_docker_environment.py:913
↓ 2 callers
Function
_skill
(path: Path)
src/skillevaluator/tier3/decoys.py:110
↓ 2 callers
Function
_skill
(tmp_path: Path)
tests/tier3/test_generated_script_line_endings.py:46
↓ 2 callers
Function
_skill
(root: Path)
tests/tier3/test_installed_agent_bootstrap.py:367
↓ 2 callers
Function
_skill_entry
(path: str, fingerprint: str)
tests/deduplication/plugin/test_openclaw_profile.py:97
↓ 2 callers
Function
_skill_load
(name: str, arguments: str = "")
tests/tier3/test_f90_signals_handoff_prompt.py:48
↓ 2 callers
Method
_skill_manifest_fields
Project the source client's skill declarations once, for the manifest and the census. Cursor loads undotted roots as written, so those become
src/skillevaluator/tier3/plugin_native.py:1410
↓ 2 callers
Function
_skill_match_details
(expected_skill: str, acceptable_skills: Any = None)
src/skillevaluator/tier3/eval_core/checks.py:3275
↓ 2 callers
Function
_skill_md_arg
(arg: str, assignments: dict[str, str])
src/skillevaluator/tier3/eval_core/checks.py:1703
↓ 2 callers
Function
_skill_name_from_ref
(ref: str)
src/skillevaluator/tier3/eval_core/checks.py:5383
↓ 2 callers
Function
_skill_sizes
Size of each SKILL.md under skills/ that is a readable regular file. Pipes, sockets, devices, dangling or looping links and unreadable files are
src/skillevaluator/harness/plugin.py:59
↓ 2 callers
Function
_skill_use_assertion_re
Graded text about using the skill: its files, its name called a skill, "used <name>", or running its scripts. A skill named after the CLI or file
src/skillevaluator/tier3/evals_spec.py:744
↓ 2 callers
Function
_skillspector_documentation_only_report
Return the exact SkillSpector 2.11.2 docs-only applicability shape.
tests/validators/test_security.py:271
↓ 2 callers
Function
_skillspector_flags_truncation
Return whether a report object has a ``*_truncated`` key that is neither ``false`` nor ``null``. SkillSpector flags a list it cut short this way,
src/skillevaluator/validators/security.py:532
↓ 2 callers
Function
_skillspector_glob_matcher
Match the names of at most *longest* characters that the shell glob *pattern* may match, else ``None``. ``*`` matches any text and ``?`` one char
src/skillevaluator/validators/security.py:1984
↓ 2 callers
Function
_skillspector_missing_references_have_rows
Whether *references* holds only objects, none ``ambiguous``, each ``missing`` one on the line of a row in *rows*. Only the line of a row is read
src/skillevaluator/validators/security.py:2454
↓ 2 callers
Function
_skillspector_partial_report
Return a valid partial report whose only gaps are *ledger_exceptions*.
tests/validators/test_security.py:337
↓ 2 callers
Function
_skillspector_passive_raster_note
A note for a ``partial`` SkillSpector analysis whose only gaps are referenced images, else ``None``. SkillSpector's ``static_patterns_*`` analyze
src/skillevaluator/validators/security.py:1470
↓ 2 callers
Function
_skillspector_provider_retries_exhausted
Return whether a failed LLM-stage report shows only an LLM provider outage. Every ledger exception must be a fatal ``llm_connection_retries_exhau
src/skillevaluator/validators/security.py:544
↓ 2 callers
Function
_skillspector_provider_retry_error
Describe a provider outage in fixed text; no report field is quoted.
src/skillevaluator/validators/security.py:629
↓ 2 callers
Function
_skillspector_reference_caveat_error
Describe a ``complete_with_caveats`` analysis whose caveats were not accepted, naming up to three. When some caveats are not missing references,
src/skillevaluator/validators/security.py:2683
↓ 2 callers
Function
_skillspector_reference_caveat_notes
Warnings for a ``complete_with_caveats`` analysis, else ``None``. SkillSpector 2.13.4 and later report an analysis whose only gaps are nonfat
src/skillevaluator/validators/security.py:2641
↓ 2 callers
Function
_skillspector_reference_evidence_unmatched
Whether a report whose ledger has no missing-reference row lists a gap that nothing reads. SkillSpector gives every ``missing`` or ``ambiguous``
src/skillevaluator/validators/security.py:2472
↓ 2 callers
Function
_skillspector_references_on_line
The entries of a report's ``references`` list for line *line* of *path*.
src/skillevaluator/validators/security.py:1839
↓ 2 callers
Function
_skillspector_row_is_reference_caveat
Whether a ledger row is a missing reference, the only caveat of a ``complete_with_caveats`` analysis.
src/skillevaluator/validators/security.py:2625
↓ 2 callers
Function
_skillspector_rows_are_reference_caveats
Whether a ``complete_with_caveats`` ledger has caveats and every one is a missing reference. SkillSpector 2.13.4 and later give that status only
src/skillevaluator/validators/security.py:2630
↓ 2 callers
Function
_skillspector_shell_patterns
Glob patterns for the path parts of the words in *texts* that a shell or Windows expands or unquotes. The words are split at whitespace and the M
src/skillevaluator/validators/security.py:2258
↓ 2 callers
Function
_skillspector_stand_in
The first code point of :data:`_SKILLSPECTOR_STAND_IN_RANGES` that no text of *texts* holds. So a stand-in for syntax in the texts being read is
src/skillevaluator/validators/security.py:2102
↓ 2 callers
Function
_skillspector_timeout_seconds
Return the SkillSpector subprocess watchdog configured by the operator.
src/skillevaluator/validators/security.py:642
↓ 2 callers
Function
_skillspector_universal_analyzers_completed
Whether each universal analyzer completed one work item per component and no other analyzer degraded.
src/skillevaluator/validators/security.py:2603
↓ 2 callers
Function
_skillspector_unmodeled_hook_manifests
The manifests whose BH1 finding says SkillSpector parsed hooks but could not model their payload. Each manifest is the BH1 location's path when i
src/skillevaluator/validators/security.py:2742
↓ 2 callers
Function
_skillspector_unscanned_paths
Relative paths of entries that SkillSpector did not inspect, else ``None``. An entry is uninspected when it is a link, a junction or another
src/skillevaluator/validators/security.py:1851
↓ 2 callers
Function
_skillspector_version
Read an unambiguous SkillSpector identity line, preserving release suffixes.
src/skillevaluator/utils/tool_runner.py:130
↓ 2 callers
Function
_skip_context
(reason, **metadata)
src/skillevaluator/deduplication/plugin/commands.py:54
↓ 2 callers
Function
_skip_js_quoted
(source: str, start: int, quote: str)
src/skillevaluator/tier3/eval_core/codex_tool_call_normalizer.py:162
↓ 2 callers
Function
_skip_newlines
(tokens: list[tuple[str, str]], cursor: int)
src/skillevaluator/tier3/eval_core/codex_tool_call_normalizer.py:610
↓ 2 callers
Function
_skip_reason
Return a bounded, terminal-safe reason that keeps its line structure. Each line is flattened and bounded separately, and at most ``_SKIP_REAS
src/skillevaluator/reporting/console_ui.py:1005
↓ 2 callers
Function
_skip_reasons
(results: list[ValidationResult])
src/skillevaluator/reporting/benchmark.py:2092
↓ 2 callers
Function
_skip_toml_string_literal
Return index immediately after the TOML string literal starting at *start*.
src/skillevaluator/tier3/harbor/adapter.py:4788
↓ 2 callers
Function
_skip_whitespace
(index: int)
src/skillevaluator/tier3/eval_core/llm_judge.py:402
↓ 2 callers
Function
_skip_whitespace
(index)
src/skillevaluator/tier3/harbor/templates/eval.py:2699
↓ 2 callers
Function
_skipped_result
(check_id: str, name: str)
src/skillevaluator/tier1/commands.py:391
↓ 2 callers
Function
_snapshot_evaluator_parent_path
Capture every real directory component anchoring an evaluator file.
src/skillevaluator/tier3/harbor/adapter.py:1781
↓ 2 callers
Function
_snapshot_ids
(result: dict[str, Any])
tests/test_tier3_max_cases.py:129
↓ 2 callers
Function
_source_digest
(path: Path)
src/skillevaluator/tier3/decoys.py:146
↓ 2 callers
Function
_source_identity
(root: Path, manifest: str)
src/skillevaluator/tier3/plugin_prepare.py:81
↓ 2 callers
Function
_source_reaches_judge
Return True when every character of *source* appears in *evidence*. A long source can arrive in two pieces: the compact history repeats the f
src/skillevaluator/tier3/eval_core/atif_helpers.py:1126
↓ 2 callers
Method
_span
(self, start, end)
tests/deduplication/utils/test_chunker.py:60
↓ 2 callers
Function
_spend_strict_judge_request
Count one request against the strict per-model limit; a no-op outside the strict policy.
src/skillevaluator/tier3/harbor/templates/eval.py:1038
↓ 2 callers
Function
_split_canonical
Split a :func:`normalize_ref` result into ``(source, repo, kind, name)`` (see :func:`parse_canonical_ref`).
src/skillevaluator/plugin_dependencies.py:118
↓ 2 callers
Function
_split_own_heading
Return the chunk's own heading line and the text after it. A Markdown section chunk opens with its heading line, so the same body under diffe
src/skillevaluator/deduplication/intra_skill/llm_analyzer.py:132
↓ 2 callers
Function
_split_raw_userinfo
``(userinfo, the URL without it)``, read from the raw text as :func:`url_credentials` reads it. The authority runs from ``//`` to the next ``/``,
src/skillevaluator/validators/url_policy.py:479
↓ 2 callers
Method
_split_results_by_tier
Bucket a flat results list into ``(tier1, tier2, tier3)`` slices. ``commands.validate`` concatenates results in tier order before handing
src/skillevaluator/reporting/html.py:2961
↓ 2 callers
Function
_ssh_to_https
Convert a git remote URL to a browsable HTTPS URL. Handles: ssh://git@host:port/group/repo.git -> https://host/group/repo git@hos
src/skillevaluator/utils/helpers.py:228
↓ 2 callers
Function
_stable_read_metadata
Select comparable metadata after a descriptor-pinned read.
src/skillevaluator/tier3/output_provenance.py:56
↓ 2 callers
Function
_stage
(results: list[ValidationResult])
src/skillevaluator/tier_workflows.py:48
↓ 2 callers
Function
_stage
(tmp_path: Path, case_id: str)
tests/tier3/test_reference_custom_grader_fixtures.py:16
↓ 2 callers
Function
_stage_arguments
The arguments a pipeline stage passes its interpreter, as the shell reads them. Quotes are removed. An unquoted ``#`` at the start of a word ends
src/skillevaluator/plugin_component_risk.py:2297
↓ 2 callers
Function
_stage_native_harbor_tasks_into
Build native Harbor tasks inside a private, caller-owned directory. The source tree is copied first and all SkillEvaluator injections happen only
src/skillevaluator/tier3/harbor/adapter.py:5816
↓ 2 callers
Method
_stage_rules
Stage each rule as a Claude Code user rule; return the setup lines that copy them into place. The rules go to ``$CLAUDE_CONFIG_DIR/rules/skil
src/skillevaluator/tier3/plugin_native.py:1726
↓ 2 callers
Function
_stage_runs_stdin
Whether a pipeline stage runs its standard input as a program ('| sh', '| python3 -', '| iex').
src/skillevaluator/plugin_component_risk.py:2255
↓ 2 callers
Function
_stage_task
(tmp_path: Path, configured: str | None, *, native: bool = False)
tests/tier3/test_security_workdir_provenance.py:17
↓ 2 callers
Function
_staged_claude_hooks
(tmp_path: Path)
tests/tier3/test_plugin_execution.py:212
↓ 2 callers
Function
_staged_handler_needle
A one-line ``"key": value`` fragment of the source's first non-``type`` handler field. The staged ``hooks.json`` is indent-2 JSON, so each scalar
src/skillevaluator/tier3/plugin_native.py:1304
↓ 2 callers
Function
_staged_hook_events
The part of a hook source Claude Code loads: a Cursor source keeps only translatable events.
src/skillevaluator/tier3/plugin_native.py:1327
↓ 2 callers
Function
_staged_modules
(tmp_path: Path)
tests/tier3/test_verifier_python39_syntax.py:16
↓ 2 callers
Function
_staged_rules
Split the plugin rules into always-on ``(name, body)`` pairs and census ``not_loaded`` rows.
src/skillevaluator/tier3/plugin_native.py:1228
↓ 2 callers
Function
_staged_task_environment_error
Reject staged task environments the selected backend cannot run.
src/skillevaluator/tier3/harbor/runner.py:1806
↓ 2 callers
Function
_stamp_tier1_scan_scope
Record certified source paths that the Tier 1 scans skipped by name.
src/skillevaluator/tier1/commands.py:168
↓ 2 callers
Function
_starts_absolute_path
Whether a POSIX, home, UNC, drive-letter or home-variable absolute path starts at ``index``. The character before it must not belong to a name, s
src/skillevaluator/evaluation/tier3_feedback_chat.py:476
↓ 2 callers
Method
_static_test_evidence_message
Return the static test limitation from direct or folder-aggregated results.
src/skillevaluator/reporting/cli.py:953
↓ 2 callers
Function
_status
(result: dict)
ci/report_views.py:65
↓ 2 callers
Function
_statuses
(proof: dict[str, Any])
tests/tier3/test_f90_mcp_proof_status.py:42
↓ 2 callers
Method
_stop
()
src/skillevaluator/reporting/assets/chart.umd.min.js:13
↓ 2 callers
Method
_stop_live
Stop Live, leaving its last frame on lines of its own.
src/skillevaluator/tier3/harbor/progress.py:551
↓ 2 callers
Function
_stored_key_size_is_valid
(size: int)
src/skillevaluator/tier3/output_provenance.py:328
↓ 2 callers
Function
_stored_tools
The bounded copy of a tool list kept in a privilege record; risk checks use the full list.
src/skillevaluator/plugin_component_risk.py:590
↓ 2 callers
Function
_strict_judge_error_stops
Return whether a provider error is a quota, budget, billing, safety or security rejection. Only what the provider said is read: the code, type an
src/skillevaluator/tier3/harbor/templates/eval.py:1067
↓ 2 callers
Function
_strict_judge_failure
Describe a failed strict judge request and whether the next candidate may answer instead. Only unavailability moves on: a transient HTTP status,
src/skillevaluator/tier3/harbor/templates/eval.py:2339
↓ 2 callers
Function
_strict_judge_retry_allowed
Return whether the strict policy lets the transport retry *error* on the same model.
src/skillevaluator/tier3/harbor/templates/eval.py:1122
↓ 2 callers
Function
_strip_arm_and_attempt_suffixes
Remove at most one dual-arm suffix and one attempt suffix regardless of ordering.
src/skillevaluator/tier3/harbor/collector.py:4589
↓ 2 callers
Function
_strip_complete_mcp_placeholders
Return the string with complete $VAR and ${VAR} placeholders removed while preserving escaped $$.
src/skillevaluator/tier3/harbor/adapter.py:2017
↓ 2 callers
Function
_strip_location
Return the first line of *message* without its absolute-path location. YAML parse errors carry context lines (a copy of the offending source), so
src/skillevaluator/precommit.py:150
↓ 2 callers
Method
_strip_possessive
(self, token: str)
src/skillevaluator/evaluation/dataset_case_removal.py:290
↓ 2 callers
Function
_structural_limit_reward
Keep only shallow identifiers when a reward payload is unsafe to traverse.
src/skillevaluator/tier3/harbor/collector.py:3920
↓ 2 callers
Function
_stub_validators
(monkeypatch: pytest.MonkeyPatch, *validator_names: str)
tests/test_artifact_applicability.py:113
↓ 2 callers
Function
_style_block_digests
()
tests/reporting/test_report_redesign.py:486
↓ 2 callers
Function
_style_finding
The finding for a declared path a client does not accept as written (no leading ``./``, or ``./`` alone). Its severity follows the client: Claude
src/skillevaluator/plugin_paths.py:454
← previous
next →
4,601–4,700 of 24,741, ranked by callers