Back to skill

Security audit

lance-format

Security checks for vulnerabilities and agentic risk

Overview

This Lance reference skill is purpose-aligned, though some storage and data-writing examples need care before use.

Use fresh dataset paths for examples, HTTPS for remote storage, and credentials limited to the intended dataset. Preserve required retention protections and immutable backups; use a suitable separate bucket rather than blindly disabling Object Lock. The scanner's partial areas were independently investigated, but external dependencies and upstream provenance remain unverified.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • Rogue AgentSelf-Modification, Session Persistence
Findings (38)

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 37)May include surrounding context.

md
| `changelog-v7-v15.md` | The full v7 -> v15 delta | 14 |

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 189)May include surrounding context.

md
Full per-tag deltas with every PR citation: `references/changelog-v7-v15.md`.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 313)May include surrounding context.

md
recreating. Conditions for each in `references/changelog-v7-v15.md`.

Self-Modification

High
Category
Rogue Agent
Confidence
85% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · references/docs/format/table/blob.md (reported line 332)May include surrounding context.

md
Both bits remain set in subsequent versions, including after removing the
opt-in, dropping blob fields, or restoring an older snapshot. Activation does
not rewrite existing files: flagged snapshots can contain all six kinds, and
supporting readers must understand Packed and Dedicated as well as Managed.
Clients without the capability must reject flagged snapshots.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/docs/guide/object_store.md (reported line 505)May include surrounding context.

md
| `azure_object_id` / `object_id` | Object id for use with managed identity authentication. |
| `azure_msi_resource_id` / `msi_resource_id` | Msi resource id for use with managed identity authentication. |
| `azure_federated_token_file` / `federated_token_file` | File containing token for Azure AD workload identity federation. |
| `azure_use_azure_cli` / `use_azure_cli` | Use azure cli for acquiring access token. |
| `azure_disable_tagging` / `disable_tagging` | Disables tagging objects. This can be desirable if not supported by the backing store. |

## AliCloud Object Storage Service Configuration

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/docs/guide/object_store.md (reported line 636)May include surrounding context.

md
| Key | Description |
| --- | --- |
| `hf_token` | Hugging Face access token. Falls back to `HF_TOKEN` or `HUGGINGFACE_TOKEN` when omitted. |
| `hf_revision` | Repository revision, such as a commit ID, branch, or tag. Defaults to `main`. |
| `hf_download_mode` | `http` (default) or `xet`. |
| `hf_enable_resolve_cache` | `"true"` reuses resolved HTTP download URLs and XET file metadata across readers. Defaults to `"false"`. |

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/docs/guide/object_store.md (reported line 885)May include surrounding context.

text

   The command prints an `awsAccessKey` / `awsSecret` pair
   that maps to the Ozone user's S3 access credentials.

#### Connection

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
80% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · references/format-table.md (reported line 529)May include surrounding context.

md
bit-16384 activation rule (`versioning.md:51` now documents the explicit `lance.blob.managed =
"true"` opt-in that #9891 put in the code). The live splits:

- **512** - the table (`versioning.md:35`) gives reader/writer `Yes` with no caveat; the code says
  "Understood only by debug builds until support is released" (`feature_flags.rs:57-61`;
  `frag_reuse_with_stable_row_ids_enabled() = cfg!(debug_assertions)`, `:454-455`), so a release
  build refuses an FRI-on-stable-row-ID table outright.

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
80% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · references/indexes.md (reported line 562)May include surrounding context.

md
`TEXT_PREPPER` (`LowerCaser` + `AsciiFoldingFilter`, `:87-91`) on **both** the build and query
  paths (`:1150`, `:507`). FM-Index has no case handling at all - queries go straight to
  `search_string_contains(pattern.as_bytes())` (`fmindex.rs:1638`). The same needle therefore
  returns different result sets from the two indexes, with no warning.
- **FTS answers a different question.** It matches whole tokens, so it is the wrong tool for
  identifier or mid-token lookup: a needle that appears inside larger tokens matches nothing
  under FTS while NGRAM/FM find it, and a needle that is also a common token can match orders of

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
80% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · references/performance.md (reported line 542)May include surrounding context.

md
source schema is a strict subset of the target's and unmatched rows are kept, the builder
  routes to `RewriteColumns` (a cheap column update) rather than rewriting whole fragments
  (`rust/lance/src/dataset/write/merge_insert.rs:2232-2260`). Passing a full-schema source for
  what is logically a two-column update therefore costs dramatically more, with no warning.
  Project the source down to the key plus the columns actually being updated.
- **Local-filesystem durability is delegated, not added - by default.** Lance's `file://` store is
  `object_store::LocalFileSystem` (`rust/lance-io/src/object_store/providers/local.rs:14,25`,

Self-Modification

High
Category
Rogue Agent
Confidence
90% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · references/performance.md (reported line 711)May include surrounding context.

md
- **Real erasure is four steps, not a `delete`.** A delete writes a deletion vector; the bytes
  stay in the data file, in every older version, and in index files. Source-read at 12.0.0 and
  re-checked at `v14.0.0-beta.14` (a design review, not field-measured):
  1. **Rewrite the affected fragments.** Compaction only self-rewrites a fragment whose deleted
     share exceeds `materialize_deletions_threshold`, default `0.1`
     (`rust/lance/src/dataset/optimize.rs:353,931-934`), so a full-size fragment with a few
     deleted rows is skipped - set the threshold to `0.0` for an erasure pass.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/docs/format/index/scalar/minhash_lsh.md (reported line 383)May include surrounding context.

md
---------------------------------------------------------------------------------------------------------------|
| `minhash_lsh_details`           | both    | Lower-case hexadecimal encoding of the serialized `MinHashLshIndexDetails` the segment was built from. A reader decodes it and requires it to describe, field by field, the same details as the index metadata. |
| `minhash_lsh_index_version`     | both    | Decimal file layout version; This is equivalent to the version in the `IndexMetadata`                                                                                                           |
| `minhash_lsh_num_docs`          | `bands` | Decimal number of documents of the segment; equals the row count of `signatures.lance`.                                                                                                                       |
| `minhash_lsh_page_rows`         | `bands` | Decimal number of rows per logical page; positive. Writers use `4096`.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/docs/format/index/scalar/minhash_lsh.md (reported line 384)May include surrounding context.

md
ent was built from. A reader decodes it and requires it to describe, field by field, the same details as the index metadata. |
| `minhash_lsh_index_version`     | both    | Decimal file layout version; This is equivalent to the version in the `IndexMetadata`                                                                                                           |
| `minhash_lsh_num_docs`          | `bands` | Decimal number of documents of the segment; equals the row count of `signatures.lance`.                                                                                                                       |
| `minhash_lsh_page_rows`         | `bands` | Decimal number of rows per logical page; positive. Writers use `4096`.                                                                                                                                        |
| `minhash_lsh_page_table_buffer` | `bands` | Decimal index of the global buffer of `bands.lance` that holds the page table.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/docs/format/index/scalar/minhash_lsh.md (reported line 385)May include surrounding context.

md
dex_version`     | both    | Decimal file layout version; This is equivalent to the version in the `IndexMetadata`                                                                                                           |
| `minhash_lsh_num_docs`          | `bands` | Decimal number of documents of the segment; equals the row count of `signatures.lance`.                                                                                                                       |
| `minhash_lsh_page_rows`         | `bands` | Decimal number of rows per logical page; positive. Writers use `4096`.                                                                                                                                        |
| `minhash_lsh_page_table_buffer` | `bands` | Decimal index of the global buffer of `bands.lance` that holds the page table.                                                                                                                                |

### Page Table

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/docs/format/index/scalar/minhash_lsh.md (reported line 386)May include surrounding context.

md
ex_version`     | both    | Decimal file layout version; This is equivalent to the version in the `IndexMetadata`                                                                                                           |
| `minhash_lsh_num_docs`          | `bands` | Decimal number of documents of the segment; equals the row count of `signatures.lance`.                                                                                                                       |
| `minhash_lsh_page_rows`         | `bands` | Decimal number of rows per logical page; positive. Writers use `4096`.                                                                                                                                        |
| `minhash_lsh_page_table_buffer` | `bands` | Decimal index of the global buffer of `bands.lance` that holds the page table.                                                                                                                                |

### Page Table

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/docs/format/index/scalar/rtree.md (reported line 77)May include surrounding context.

md
-|:---------|:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `bbox` | RectType | false    | Type is Rect defined by [geoarrow-rs](https://github.com/geoarrow/geoarrow-rs) RectType; physical storage is Struct<xmin: Float64, ymin: Float64, xmax: Float64, ymax: Float64>. Represents the node bounding box (leaf: item bbox; branch: child aggregation). |
| `id`   | UInt64   | false    | Reuse the `id` column to store a row address in leaf pages and `pageid` in branch pages. A segment persisted before format version 1 (`IndexMetadata::index_version`) stores a row id in leaf pages instead.                                                                                                                                                                               |

### Nulls File Schema

Context Window Stuffing

Medium
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · references/docs/format/table/table_config.md (reported line 30)May include surrounding context.

md
<div class="table-config-table" markdown="1">

| Key                                            | Type                                                                     | Description                                                                                                                                                                                                                                                                             |
|-------------------------------------------------|---------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `lance.auto_cleanup.interval`                   | Non-negative integer                                                     | Numbe

Context Window Stuffing

Medium
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · references/docs/guide/object_store.md (reported line 62)May include surrounding context.

md
These options apply to all object stores.

| Key                          | Description                                                                                                                                                                                                                                                                                             |
|------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `allow_http`                 | Allow non-TLS, i.e. non-HTTPS connections. Default, `False`.

Context Window Stuffing

Medium
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · references/docs/guide/object_store.md (reported line 70)May include surrounding context.

md
|
| `timeout`                    | Timeout for the entire request, from connection until the response body has finished. Default, `30s`. This applies to each individual request, so on a large write it must cover one complete multipart part upload; raise it alongside `LANCE_INITIAL_UPLOAD_SIZE`.                                    |
| `user_agent`                 | User agent string to use in requests.                                                                                                                                                                                                                                                                   |
| `proxy_url`                  | URL of a proxy server to use for requests. Default, `None`.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/docs/format/table/table_config.md (reported line 30)May include surrounding context.

md
<div class="table-config-table" markdown="1">

| Key                                            | Type                                                                     | Description                                                                                                                                                                                                                                                                             |
|-------------------------------------------------|---------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `lance.auto_cleanup.interval`                   | Non-negative integer                                                     | Numbe

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/docs/format/table/table_config.md (reported line 33)May include surrounding context.

md
eanup runs on the commit whose version is a multiple of this interval. If unset, or `0`, automatic cleanup is disabled. See [Automatic Cleanup](../../guide/read_and_write.md#automatic-cleanup).                                |
| `lance.auto_cleanup.older_than`                 | Duration string (e.g. `3600s`, `2h`)                                     | Only versions older than this duration are removed by automatic cleanup. See [Automatic Cleanup](../../guide/read_and_write.md#automatic-cleanup).                                                                                                                                       |
| `lance.auto_cleanup.retain_versions`             | Non-negative integer                                                     | Minimum number of most-recent versions to retain, regardless of age, when automatic cleanup runs.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/docs/format/table/table_config.md (reported line 34)May include surrounding context.

md
sions older than this duration are removed by automatic cleanup. See [Automatic Cleanup](../../guide/read_and_write.md#automatic-cleanup).                                                                                                                                       |
| `lance.auto_cleanup.retain_versions`             | Non-negative integer                                                     | Minimum number of most-recent versions to retain, regardless of age, when automatic cleanup runs.                                                                                                                                                                                        |
| `lance.auto_cleanup.referenced_branch`          | Boolean (`true`/`false`)                                                 | Whether automatic cleanup also removes eligible old versions that are only reachable through a [branch or tag](branch_tag.md) reference. Defaults to `false`.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/docs/format/table/table_config.md (reported line 35)May include surrounding context.

md
, when automatic cleanup runs.                                                                                                                                                                                        |
| `lance.auto_cleanup.referenced_branch`          | Boolean (`true`/`false`)                                                 | Whether automatic cleanup also removes eligible old versions that are only reachable through a [branch or tag](branch_tag.md) reference. Defaults to `false`.                                                                                                                            |
| `lance.auto_cleanup.delete_rate_limit`          | Non-negative integer                                                     | Maximum number of delete operations per second automatic cleanup is allowed to issue against the underlying object store.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/docs/format/table/table_config.md (reported line 42)May include surrounding context.

md
tions are materialized. `0` (or lower) materializes deletions for all fragments with deletions; values above `1.0` never materialize deletions. Defaults to `0.1`.                                                   |
| `lance.compaction.defer_index_remap`             | Boolean (`true`/`false`)                                                 | If `true`, indices are not remapped as part of this compaction. Instead, the fragment reuse index is updated, and remapping is deferred to a later operation.                                                                                                                            |
| `lance.compaction.index_remap_mode`              | `direct` or `merged`                                                     | How the old-to-new row-address mapping used to remap indices is built. Defaults to `direct`.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/docs/format/table/table_config.md (reported line 36)May include surrounding context.

md
removes eligible old versions that are only reachable through a [branch or tag](branch_tag.md) reference. Defaults to `false`.                                                                                                                            |
| `lance.auto_cleanup.delete_rate_limit`          | Non-negative integer                                                     | Maximum number of delete operations per second automatic cleanup is allowed to issue against the underlying object store.                                                                                                                                                                |
| `lance.compaction.target_rows_per_fragment`      | Non-negative integer                                                     | Target number of rows per file used to decide which fragments are candidates for [compaction](../../guide/read_and_write.md#compact-data-files). Fragments with fewer rows than this are compacted. Defaults to 1,