{"items":[{"id":"NVI-00014045","kind":"vulnerability","title":"CodeWhale: rlm_eval auto-approves arbitrary Python execution, bypassing the user's approval policy (RCE)","revision":3,"unknowns":["Customer inventory, deployment, reachability and exploitation have not been inspected.","NVD lastModified does not establish freshness of every associated CPE record."],"lifecycle":"active","identifiers":["GHSA-wrj3-vj8c-784f","CVE-2026-75858"],"plain_summary":"GitHub lists this reviewed advisory for deepseek-tui, deepseek-tui, codewhale-tui, codewhale. Published version ranges and source prerequisites are shown below; this is not an inspection of a visitor’s application.","review_status":"source_listed","control_topics":[],"public_actions":[{"summary":"Identify the exact component and deployed version, examine the linked publisher’s conditions, and retain evidence of any mitigation.","source_ids":["sr-1c0e71dea93467d72ff84d844723de68"],"action_type":"investigate","limitations":"Confirm current vendor support, compatibility and deployment evidence. This report has not tested a fix in your system.","current_supported_target":null}],"schema_version":"1.1.0","source_metrics":[{"kind":"CVSS","as_of":"2026-09-04T18:12:16.000Z","value":"7.8 high","meaning":"GitHub source severity; exploitability and local consequence require separate evidence.","source_id":"sr-37e6751a76917a4fdf587bd1cbab6d38"},{"kind":"CVSS","as_of":"2026-10-22T22:10:00.250Z","value":"7.8 HIGH (3.1)","meaning":"disclosure@vulncheck.com severity; no combined Null risk score.","source_id":"sr-27f2704e79165d1b68c85f881dd46a4c"},{"kind":"CVSS","as_of":"2026-10-22T22:10:00.250Z","value":"8.5 HIGH (4.0)","meaning":"disclosure@vulncheck.com severity; no combined Null risk score.","source_id":"sr-27f2704e79165d1b68c85f881dd46a4c"},{"kind":"EPSS","as_of":"2026-11-04T12:03:12.000Z","value":"0.364%","meaning":"FIRST EPSS model v2026.06.15; exploitation probability over 30 days, percentile 0.30449. Missing scores are unknown.","source_id":"sr-1c0e71dea93467d72ff84d844723de68"}],"why_it_matters":"Published affected context helps an owner decide what evidence to gather. This report has not examined your application.","related_sources":[{"id":"sr-37e6751a76917a4fdf587bd1cbab6d38","url":"https://github.com/advisories/GHSA-wrj3-vj8c-784f","basis":"vendor_advisory","notes":"GitHub Advisory Database; CC-BY-4.0. Each advisory links its original attribution. Null normalizes fields.","title":"GHSA-wrj3-vj8c-784f","publisher":"GitHub Advisory Database","observed_at":"2026-11-02T02:20:52.377Z"},{"id":"sr-27f2704e79165d1b68c85f881dd46a4c","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-75858","basis":"official_catalog","notes":"This product uses data from the NVD API but is not endorsed or certified by the NVD.","title":"CVE-2026-75858","publisher":"NIST National Vulnerability Database","observed_at":"2026-11-08T19:47:48.947Z"},{"id":"sr-1c0e71dea93467d72ff84d844723de68","url":"https://www.first.org/epss/data.html","basis":"official_catalog","notes":"FIRST EPSS; probability of exploitation in the next 30 days, not local compromise. Dataset date/model retained; FIRST services terms apply.","title":"CVE-2026-75858","publisher":"FIRST EPSS","observed_at":"2026-11-04T23:33:33.066Z"}],"affected_context":[{"certainty":"source_listed","source_ids":["sr-37e6751a76917a4fdf587bd1cbab6d38"],"technology":"deepseek-tui","known_exclusions":[],"conditions_summary":"### Maintainer resolution\n\nThe CodeWhale maintainers validated this report. The affected package ranges are recorded in the advisory metadata. Version 0.8.64 contains the fix in commit 57f3c89471e27ac4032d9791f6885e5d4408c381. Users should upgrade to 0.8.64 or later. The original reporter analysis is preserved below.\n\n### Summary\nThe `rlm_eval` tool runs an arbitrary Python string chosen by the model in a real `python3` interpreter. Its `approval_requirement()` returns `ApprovalRequirement::Auto`, which the engine treats as \"never prompt,\" regardless of the user's configured `--approval-policy`. A single tool call — which prompt injection from any untrusted content the agent reads (a web page, a fetched URL, a repo file, an MCP tool result) can induce — runs code on the user's machine at the user's privilege with no prompt and no audit step. This is the same defect that was already patched on the sibling `run_tests` tool (CVE-2026-45311); the fix never reached `rlm_eval` or `rlm_open`, which expose a broader surface (full Python on the host, not just `cargo test`).\n\n### Details\n`rlm_eval`'s `execute()` reads the LLM-controlled `code` field and runs it (`crates/tui/src/tools/rlm.rs:","public_affected_ranges":[">= 0.8.33, <= 0.8.41"]},{"certainty":"source_listed","source_ids":["sr-37e6751a76917a4fdf587bd1cbab6d38"],"technology":"deepseek-tui","known_exclusions":[],"conditions_summary":"### Maintainer resolution\n\nThe CodeWhale maintainers validated this report. The affected package ranges are recorded in the advisory metadata. Version 0.8.64 contains the fix in commit 57f3c89471e27ac4032d9791f6885e5d4408c381. Users should upgrade to 0.8.64 or later. The original reporter analysis is preserved below.\n\n### Summary\nThe `rlm_eval` tool runs an arbitrary Python string chosen by the model in a real `python3` interpreter. Its `approval_requirement()` returns `ApprovalRequirement::Auto`, which the engine treats as \"never prompt,\" regardless of the user's configured `--approval-policy`. A single tool call — which prompt injection from any untrusted content the agent reads (a web page, a fetched URL, a repo file, an MCP tool result) can induce — runs code on the user's machine at the user's privilege with no prompt and no audit step. This is the same defect that was already patched on the sibling `run_tests` tool (CVE-2026-45311); the fix never reached `rlm_eval` or `rlm_open`, which expose a broader surface (full Python on the host, not just `cargo test`).\n\n### Details\n`rlm_eval`'s `execute()` reads the LLM-controlled `code` field and runs it (`crates/tui/src/tools/rlm.rs:","public_affected_ranges":[">= 0.8.33, < 0.8.41"]},{"certainty":"source_listed","source_ids":["sr-37e6751a76917a4fdf587bd1cbab6d38"],"technology":"codewhale-tui","known_exclusions":[],"conditions_summary":"### Maintainer resolution\n\nThe CodeWhale maintainers validated this report. The affected package ranges are recorded in the advisory metadata. Version 0.8.64 contains the fix in commit 57f3c89471e27ac4032d9791f6885e5d4408c381. Users should upgrade to 0.8.64 or later. The original reporter analysis is preserved below.\n\n### Summary\nThe `rlm_eval` tool runs an arbitrary Python string chosen by the model in a real `python3` interpreter. Its `approval_requirement()` returns `ApprovalRequirement::Auto`, which the engine treats as \"never prompt,\" regardless of the user's configured `--approval-policy`. A single tool call — which prompt injection from any untrusted content the agent reads (a web page, a fetched URL, a repo file, an MCP tool result) can induce — runs code on the user's machine at the user's privilege with no prompt and no audit step. This is the same defect that was already patched on the sibling `run_tests` tool (CVE-2026-45311); the fix never reached `rlm_eval` or `rlm_open`, which expose a broader surface (full Python on the host, not just `cargo test`).\n\n### Details\n`rlm_eval`'s `execute()` reads the LLM-controlled `code` field and runs it (`crates/tui/src/tools/rlm.rs:","public_affected_ranges":[">= 0.8.41, < 0.8.64"]},{"certainty":"source_listed","source_ids":["sr-37e6751a76917a4fdf587bd1cbab6d38"],"technology":"codewhale","known_exclusions":[],"conditions_summary":"### Maintainer resolution\n\nThe CodeWhale maintainers validated this report. The affected package ranges are recorded in the advisory metadata. Version 0.8.64 contains the fix in commit 57f3c89471e27ac4032d9791f6885e5d4408c381. Users should upgrade to 0.8.64 or later. The original reporter analysis is preserved below.\n\n### Summary\nThe `rlm_eval` tool runs an arbitrary Python string chosen by the model in a real `python3` interpreter. Its `approval_requirement()` returns `ApprovalRequirement::Auto`, which the engine treats as \"never prompt,\" regardless of the user's configured `--approval-policy`. A single tool call — which prompt injection from any untrusted content the agent reads (a web page, a fetched URL, a repo file, an MCP tool result) can induce — runs code on the user's machine at the user's privilege with no prompt and no audit step. This is the same defect that was already patched on the sibling `run_tests` tool (CVE-2026-45311); the fix never reached `rlm_eval` or `rlm_open`, which expose a broader surface (full Python on the host, not just `cargo test`).\n\n### Details\n`rlm_eval`'s `execute()` reads the LLM-controlled `code` field and runs it (`crates/tui/src/tools/rlm.rs:","public_affected_ranges":[">= 0.8.41, < 0.8.64"]}],"null_published_at":"2026-11-08T19:53:44.042Z","source_updated_at":"2026-09-04T18:12:16.000Z","source_disclosed_at":"2026-09-04T18:12:15.000Z","is_synthetic_fixture":false,"null_last_reviewed_at":null,"public_change_summary":"Source facts changed; compare the preserved prior revision.","null_first_observed_at":"2026-11-02T02:20:52.377Z"},{"id":"NVI-00014045","kind":"vulnerability","title":"CodeWhale: rlm_eval auto-approves arbitrary Python execution, bypassing the user's approval policy (RCE)","revision":1,"unknowns":["Customer inventory, deployment, reachability and exploitation have not been inspected."],"lifecycle":"active","identifiers":["GHSA-wrj3-vj8c-784f","CVE-2026-75858"],"plain_summary":"GitHub lists this reviewed advisory for deepseek-tui, deepseek-tui, codewhale-tui, codewhale. Published version ranges and source prerequisites are shown below; this is not an inspection of a visitor’s application.","review_status":"source_listed","control_topics":[],"public_actions":[{"summary":"Identify the exact component and deployed version, examine the linked publisher’s conditions, and retain evidence of any mitigation.","source_ids":["sr-37e6751a76917a4fdf587bd1cbab6d38"],"action_type":"investigate","limitations":"Confirm current vendor support, compatibility and deployment evidence. This report has not tested a fix in your system.","current_supported_target":null}],"schema_version":"1.1.0","source_metrics":[{"kind":"CVSS","as_of":"2026-09-04T18:12:16.000Z","value":"7.8 high","meaning":"GitHub source severity; exploitability and local consequence require separate evidence.","source_id":"sr-37e6751a76917a4fdf587bd1cbab6d38"}],"why_it_matters":"Published affected context helps an owner decide what evidence to gather. This report has not examined your application.","related_sources":[{"id":"sr-37e6751a76917a4fdf587bd1cbab6d38","url":"https://github.com/advisories/GHSA-wrj3-vj8c-784f","basis":"vendor_advisory","notes":"GitHub Advisory Database; CC-BY-4.0. Each advisory links its original attribution. Null normalizes fields.","title":"GHSA-wrj3-vj8c-784f","publisher":"GitHub Advisory Database","observed_at":"2026-11-02T02:20:52.377Z"}],"affected_context":[{"certainty":"source_listed","source_ids":["sr-37e6751a76917a4fdf587bd1cbab6d38"],"technology":"deepseek-tui","known_exclusions":[],"conditions_summary":"### Maintainer resolution\n\nThe CodeWhale maintainers validated this report. The affected package ranges are recorded in the advisory metadata. Version 0.8.64 contains the fix in commit 57f3c89471e27ac4032d9791f6885e5d4408c381. Users should upgrade to 0.8.64 or later. The original reporter analysis is preserved below.\n\n### Summary\nThe `rlm_eval` tool runs an arbitrary Python string chosen by the model in a real `python3` interpreter. Its `approval_requirement()` returns `ApprovalRequirement::Auto`, which the engine treats as \"never prompt,\" regardless of the user's configured `--approval-policy`. A single tool call — which prompt injection from any untrusted content the agent reads (a web page, a fetched URL, a repo file, an MCP tool result) can induce — runs code on the user's machine at the user's privilege with no prompt and no audit step. This is the same defect that was already patched on the sibling `run_tests` tool (CVE-2026-45311); the fix never reached `rlm_eval` or `rlm_open`, which expose a broader surface (full Python on the host, not just `cargo test`).\n\n### Details\n`rlm_eval`'s `execute()` reads the LLM-controlled `code` field and runs it (`crates/tui/src/tools/rlm.rs:","public_affected_ranges":[">= 0.8.33, <= 0.8.41"]},{"certainty":"source_listed","source_ids":["sr-37e6751a76917a4fdf587bd1cbab6d38"],"technology":"deepseek-tui","known_exclusions":[],"conditions_summary":"### Maintainer resolution\n\nThe CodeWhale maintainers validated this report. The affected package ranges are recorded in the advisory metadata. Version 0.8.64 contains the fix in commit 57f3c89471e27ac4032d9791f6885e5d4408c381. Users should upgrade to 0.8.64 or later. The original reporter analysis is preserved below.\n\n### Summary\nThe `rlm_eval` tool runs an arbitrary Python string chosen by the model in a real `python3` interpreter. Its `approval_requirement()` returns `ApprovalRequirement::Auto`, which the engine treats as \"never prompt,\" regardless of the user's configured `--approval-policy`. A single tool call — which prompt injection from any untrusted content the agent reads (a web page, a fetched URL, a repo file, an MCP tool result) can induce — runs code on the user's machine at the user's privilege with no prompt and no audit step. This is the same defect that was already patched on the sibling `run_tests` tool (CVE-2026-45311); the fix never reached `rlm_eval` or `rlm_open`, which expose a broader surface (full Python on the host, not just `cargo test`).\n\n### Details\n`rlm_eval`'s `execute()` reads the LLM-controlled `code` field and runs it (`crates/tui/src/tools/rlm.rs:","public_affected_ranges":[">= 0.8.33, < 0.8.41"]},{"certainty":"source_listed","source_ids":["sr-37e6751a76917a4fdf587bd1cbab6d38"],"technology":"codewhale-tui","known_exclusions":[],"conditions_summary":"### Maintainer resolution\n\nThe CodeWhale maintainers validated this report. The affected package ranges are recorded in the advisory metadata. Version 0.8.64 contains the fix in commit 57f3c89471e27ac4032d9791f6885e5d4408c381. Users should upgrade to 0.8.64 or later. The original reporter analysis is preserved below.\n\n### Summary\nThe `rlm_eval` tool runs an arbitrary Python string chosen by the model in a real `python3` interpreter. Its `approval_requirement()` returns `ApprovalRequirement::Auto`, which the engine treats as \"never prompt,\" regardless of the user's configured `--approval-policy`. A single tool call — which prompt injection from any untrusted content the agent reads (a web page, a fetched URL, a repo file, an MCP tool result) can induce — runs code on the user's machine at the user's privilege with no prompt and no audit step. This is the same defect that was already patched on the sibling `run_tests` tool (CVE-2026-45311); the fix never reached `rlm_eval` or `rlm_open`, which expose a broader surface (full Python on the host, not just `cargo test`).\n\n### Details\n`rlm_eval`'s `execute()` reads the LLM-controlled `code` field and runs it (`crates/tui/src/tools/rlm.rs:","public_affected_ranges":[">= 0.8.41, < 0.8.64"]},{"certainty":"source_listed","source_ids":["sr-37e6751a76917a4fdf587bd1cbab6d38"],"technology":"codewhale","known_exclusions":[],"conditions_summary":"### Maintainer resolution\n\nThe CodeWhale maintainers validated this report. The affected package ranges are recorded in the advisory metadata. Version 0.8.64 contains the fix in commit 57f3c89471e27ac4032d9791f6885e5d4408c381. Users should upgrade to 0.8.64 or later. The original reporter analysis is preserved below.\n\n### Summary\nThe `rlm_eval` tool runs an arbitrary Python string chosen by the model in a real `python3` interpreter. Its `approval_requirement()` returns `ApprovalRequirement::Auto`, which the engine treats as \"never prompt,\" regardless of the user's configured `--approval-policy`. A single tool call — which prompt injection from any untrusted content the agent reads (a web page, a fetched URL, a repo file, an MCP tool result) can induce — runs code on the user's machine at the user's privilege with no prompt and no audit step. This is the same defect that was already patched on the sibling `run_tests` tool (CVE-2026-45311); the fix never reached `rlm_eval` or `rlm_open`, which expose a broader surface (full Python on the host, not just `cargo test`).\n\n### Details\n`rlm_eval`'s `execute()` reads the LLM-controlled `code` field and runs it (`crates/tui/src/tools/rlm.rs:","public_affected_ranges":[">= 0.8.41, < 0.8.64"]}],"null_published_at":"2026-11-02T10:00:37.200Z","source_updated_at":"2026-09-04T18:12:16.000Z","source_disclosed_at":"2026-09-04T18:12:15.000Z","is_synthetic_fixture":false,"null_last_reviewed_at":null,"public_change_summary":"First indexed by Null; source disclosure date is shown separately.","null_first_observed_at":"2026-11-02T02:20:52.377Z"}],"next_cursor":null,"snapshot_revision":"42594"}