Skip to content
skillsafebeta
← Report library
Purpose / Documents

Firecrawl Knowledge Base Skill Security Audit

What the author says it does (original text)

Build a knowledge base from web content with Firecrawl. Use for local reference docs, RAG-ready chunks, fine-tuning datasets, documentation mirrors, topic corpora, or LLM-ready markdown organized from web sources.

Independent security check

Do not install or run it yet

Files checked
1
Risks found
4
Could it run dangerous commands?Looks for programs run straight after downloading, remote control of your computer, and hidden commands.No risks found
Could it expose your files or keys?Looks for uploads of files containing passwords or keys, and keys written directly in the code.Risks found: 1
Medium risk

Internal URLs or URLs containing sensitive parameters may be sent to hosted Firecrawl

Source references: 3
What we found

The API key is described as being for “hosted Firecrawl requests,” while source URLs are mapped and scraped. Thus, supplied URLs and accessible content may be processed by a third party, without a required check that the source is public and non-sensitive.

Why this matters

Internal hostnames, unpublished documents, tokens in query strings, or restricted page content could be exposed to the provider and become subject to its logging, retention, and access policies.

The Skill explicitly requires a key for hosted Firecrawl requests and directs Firecrawl to map and scrape sources, so URLs supplied to the workflow would at least be sent as part of hosted requests, with fetched results processed by that service. “Internal or sensitive URLs” is only a conditional scenario: the source does not show intranet access or bypass of access controls. If URLs, query parameters, or page contents are sensitive, third-party processing creates a disclosure risk. Users can ask about retention and allow only public, sanitized URLs.

SKILL.md:11In the instructionsOpen original file
inputs:  - name: FIRECRAWL_API_KEY    description: Firecrawl API key for hosted Firecrawl requests.    required: true---
Show 2 other places
SKILL.md:28In the instructionsOpen original file
Use Firecrawl map for documentation sites, search for topic-based corpora, scrape pages into markdown, and preserve code examples and tables.
SKILL.md:22In the instructionsOpen original file
Infer the source, goal, depth, and output location from context. If the source and goal are clear, proceed immediately.Ask at most 1-3 concise questions only if blocked, such as the source URL/topic, whether the output is reference/RAG/training/docs, or training format if training is requested.
Could it delete files or keep running?Looks for broad file deletion, disk overwrites, and programs set to start automatically.No risks found
Could it bypass safety checks?Looks for skipped website security checks, excessive file access, or actions that skip your approval.No risks found
Could it mislead the AI or hide text?Checks the skill instructions for requests to ignore you, influence the report, or hide text in invisible characters.Risks found: 1
High risk

Scraped content can enter RAG, training data, or agent context without instruction-safety handling

Source references: 4
What we found

The Skill scrapes web pages into Markdown and explicitly produces RAG chunks, training data, and material for agent context, but it does not require untrusted prompts or tool-use inducements embedded in pages to be removed or isolated.

Why this matters

A malicious or compromised page could plant instructions in the knowledge base. A downstream model retrieving or learning from them could be induced to expose data, invoke tools, or make incorrect decisions.

This is an active workflow, not an example or warning: it scrapes pages into Markdown and supports RAG chunks, training data, and agent context. If a page contains prompt injections, fake instructions, or tool-use lures, that text could enter later model inputs. The visible quality controls mention navigation cleanup but not treating scraped text as untrusted or filtering instructions. Users can ask for content isolation, prompt-injection checks, and human review, and restrict use in agent context or training sets.

SKILL.md:28In the instructionsOpen original file
Use Firecrawl map for documentation sites, search for topic-based corpora, scrape pages into markdown, and preserve code examples and tables.
Show 3 other places
SKILL.md:49In the instructionsOpen original file
- Reference: markdown files, `index.md`, and `sources.json`.- RAG: markdown files plus chunk files and `manifest.json`.- Training: scraped source files plus `training-data.jsonl` and `training-metadata.json`.- Docs mirror: complete markdown mirror with a table of contents.
SKILL.md:68In the instructionsOpen original file
## Usage Notes[How to use in RAG, docs, training, or agent context]
SKILL.md:84In the instructionsOpen original file
- Preserve code examples and formatting.- Remove boilerplate navigation where possible.- Include source URLs in frontmatter or metadata.
Could it change links or payment recipients without asking?Looks for forced referral or payment changes combined with instructions to hide the change.Risks found: 2
Medium risk

Inferred crawl depth and parallel execution can consume API quota without cost confirmation

Source references: 3
What we found

The Skill tells the agent to infer depth and proceed immediately when the goal appears clear, while also suggesting parallel work across sources or stages. It sets no page, request, concurrency, or spending limit and requires no cost confirmation.

Why this matters

A large documentation site, broad topic, or “exhaustive” run could rapidly consume Firecrawl quota or incur charges and generate much more data than expected.

The instructions tell the agent to infer depth and proceed immediately when source and goal appear clear, while permitting parallel work by documentation section, source type, or processing stage. With a hosted service billed or limited by requests, this could expand usage before the user confirms page count or budget. Parallelism is conditional, and the source does not prove charges will occur, but it states no request, page, concurrency, or spending cap. Users can require an estimate and hard limits before collection begins.

SKILL.md:22In the instructionsOpen original file
Infer the source, goal, depth, and output location from context. If the source and goal are clear, proceed immediately.Ask at most 1-3 concise questions only if blocked, such as the source URL/topic, whether the output is reference/RAG/training/docs, or training format if training is requested.
Show 2 other places
SKILL.md:41In the instructionsOpen original file
If appropriate, use sub-agents or equivalent parallel task runners:- one docs section per researcher- official docs, tutorials, community discussions, and references by source type- source scraping vs chunk generation vs manifest generation
SKILL.md:74In the instructionsOpen original file
## Rerun Inputsworkflow: firecrawl-knowledge-basesource: [url/topic]goal: [reference/rag/train/docs]depth: [quick/thorough/exhaustive]output_dir: [.firecrawl/]```
Medium risk

Arbitrary web content can be copied into mirrors or training sets without a rights check

Source references: 3
What we found

The Skill supports topic searches, preservation of source material, complete documentation mirrors, and training datasets. Its visible instructions require source URLs but do not require verification of licenses, site terms, or training rights.

Why this matters

Copying and reusing copyright- or contract-restricted material—especially for model training or a complete mirror—could lead to takedowns, account disputes, or legal and commercial liability. Attribution alone is not permission.

This is active functionality: the Skill can collect topic corpora, preserve code and tables, and produce complete documentation mirrors or training datasets. Recording source URLs helps attribution and traceability but does not establish permission to copy, redistribute, or train on the material; no license or site-terms check is visible. If a source is subject to copyright, contract, or crawling restrictions, mirroring or training use may expose the user to legal or account risk. Users can require allowlisted sources with recorded licenses, terms, and exclusions.

SKILL.md:28In the instructionsOpen original file
Use Firecrawl map for documentation sites, search for topic-based corpora, scrape pages into markdown, and preserve code examples and tables.
Show 2 other places
SKILL.md:51In the instructionsOpen original file
- RAG: markdown files plus chunk files and `manifest.json`.- Training: scraped source files plus `training-data.jsonl` and `training-metadata.json`.- Docs mirror: complete markdown mirror with a table of contents.
SKILL.md:84In the instructionsOpen original file
- Preserve code examples and formatting.- Remove boilerplate navigation where possible.- Include source URLs in frontmatter or metadata.

Inside this skill

6 instruction sections

The Skill directs an agent to use Firecrawl's hosted service to map documentation sites or search by topic, then scrape pages into Markdown while preserving code and tables.

View source
SKILL.md:11In the instructionsOpen original file
inputs:  - name: FIRECRAWL_API_KEY    description: Firecrawl API key for hosted Firecrawl requests.    required: true---
SKILL.md:28In the instructionsOpen original file
Use Firecrawl map for documentation sites, search for topic-based corpora, scrape pages into markdown, and preserve code examples and tables.

Outputs are written under a local `.firecrawl/` hierarchy and may include a web-content mirror, RAG chunks and manifests, or training datasets.

View source
SKILL.md:30In the instructionsOpen original file
For files, follow the Firecrawl download-style convention:```text.firecrawl/  <hostname>/    <path>/      index.md```
SKILL.md:49In the instructionsOpen original file
- Reference: markdown files, `index.md`, and `sources.json`.- RAG: markdown files plus chunk files and `manifest.json`.- Training: scraped source files plus `training-data.jsonl` and `training-metadata.json`.- Docs mirror: complete markdown mirror with a table of contents.

When the source and goal appear clear, the Skill tells the agent to proceed immediately and permits parallel work split by site section, source type, or processing stage.

View source
SKILL.md:22In the instructionsOpen original file
Infer the source, goal, depth, and output location from context. If the source and goal are clear, proceed immediately.Ask at most 1-3 concise questions only if blocked, such as the source URL/topic, whether the output is reference/RAG/training/docs, or training format if training is requested.
SKILL.md:41In the instructionsOpen original file
If appropriate, use sub-agents or equivalent parallel task runners:- one docs section per researcher- official docs, tutorials, community discussions, and references by source type- source scraping vs chunk generation vs manifest generation
Start here · InstructionsSKILL.md
firecrawl-knowledge-base
Lines connect the instruction file to its sections, not an observed execution order. Select a section to read the source.
Files and check records1 files

Coverage and gaps

Content covered in each file

These are the source ranges included in this check, not a guarantee that every issue has been resolved.

  • SKILL.mdFull text included

This report is for the version above. We read the available code and instructions without running the skill or checking extra packages it installs. This is not a promise of safety: a different version or setup may behave differently.

  • SKILL.mdInstructions

Operations mentioned in code and instructions

Connect to websites
SKILL.md:8In the instructionsOpen original file
  version: "0.1.0"  homepage: https://www.firecrawl.dev  source: https://github.com/firecrawl/firecrawl-workflows
SKILL.md:9In the instructionsOpen original file
  homepage: https://www.firecrawl.dev  source: https://github.com/firecrawl/firecrawl-workflowsinputs:
SKILL.md:12In the instructionsOpen original file
  - name: FIRECRAWL_API_KEY    description: Firecrawl API key for hosted Firecrawl requests.    required: true
Read keys or account settings
SKILL.md:11In the instructionsOpen original file
inputs:  - name: FIRECRAWL_API_KEY    description: Firecrawl API key for hosted Firecrawl requests.
Lines read
87
File checksum (to compare versions)
c5293c54349dcd7e539a35831885563e8bca320ac192ed353e622b459ab81667