prismor
Enrichment pendingRuntime Firewall for AI agents which catches the rogue tool call before it runs. Dangerous commands, secret leaks, prompt injection. For Claude Code, Codex and framework SDKs
GraphCanon updated today · GitHub synced today
Verify the decision
Maintenance and security
Full trust report- Maintenance
- Very active (0d since push)
- As of today
- Provenance
- Not a fork · Organization account
- As of today
- Security (OSV)
- No criticals
- As of today
Public GitHub metadata and optional OSV scans. Signals, not a guarantee. Trust methodology.
Install
pip install prismorSimilar tools
Same-category neighbours. No typed graph edges are catalogued for this tool yet.
Evidence and technical details
Sourced facts, taxonomy, compatibility claims, README excerpt, and machine-readable endpoints.
Overview
Runtime Firewall for AI agents which catches the rogue tool call before it runs. Dangerous commands, secret leaks, prompt injection. For Claude Code, Codex and framework SDKs
Capability facts
- CLI
- CLI entrypoint
Source: pyproject.toml:[project.scripts] · Jul 15, 2026
- MCP server
- No MCP server detected
Source: repo_scan · Jul 15, 2026
- Languages
- python, javascript
Source: github.language+package.json+pyproject.toml · Jul 15, 2026
Categories
Compatibility
Sourced claims from the README excerpt - not unsourced marketing copy.
Source: README excerpt (regex_v1, Jul 15, 2026)
nstall pyyaml # on Debian/Ubuntu use: sudo apt install python3-yamlSource link
Tags
README
Platform-specific Install
Option A: curl (easiest):
curl -sSL https://prismor.dev/install | sh
Detects your environment and uses the right install method automatically.
Option B: give your agent a skill (zero-interrupt setup):
Point your agent at SKILL.md. It is a standing instruction file: the agent reads it at session start, checks whether Prismor is installed, and follows the decision tree throughout the session without pausing your workflow.
For Claude Code, add to your CLAUDE.md:
Read `SKILL.md` and follow its instructions for runtime security.
Or via raw URL (works in any agent config file: CLAUDE.md, AGENTS.md, .cursorrules, .windsurfrules):
Read `https://raw.githubusercontent.com/PrismorSec/prismor/main/SKILL.md` and follow its instructions.
See SKILL.md for the full decision tree and hard rules.
Option C: pip:
pip install prismor
prismor setup # interactive 4-step onboarding wizard
prismor setup lets you pick enforcement mode, toggle detection rules, select agents, and optionally enable secret cloaking. Pass --non-interactive to skip the TUI.
Option D: git clone + wizard:
pip3 install pyyaml # on Debian/Ubuntu use: sudo apt install python3-yaml
git clone https://github.com/PrismorSec/prismor.git ~/.prismor
PRISMOR_MODE=enforce PRISMOR_CLOAK=1 bash ~/.prismor/scripts/init.sh .
If you are testing from a source checkout on a machine that already has a
different prismor install, use the repo shim for health checks:
python3 ~/.prismor/bin/prismor --version
python3 ~/.prismor/bin/prismor status
That path forces imports to resolve to the checked-out runtime instead of a
stale package earlier on sys.path.
On externally-managed Pythons (PEP 668 — Ubuntu 23.04+, Homebrew)
pip3 installrefuses to run; install PyYAML from your system package manager instead (sudo apt install python3-yaml,brew install pyyaml, …).init.shwill tell you if it's missing.
This installs enforce-mode Prismor hooks and the Cloak prevention layer. To register a secret, run prismor cloak add stripe_key and enter the value when prompted. To import an entire dotenv file at once, run prismor cloak add --env-file .env. Claude/Hermes can auto-decloak placeholders at the tool boundary. Codex hooks are block-only, so run placeholder commands through prismor cloak run -- <command>.
Prefer the interactive wizard? Drop the env vars:
bash ~/.prismor/scripts/init.sh .
For agents
This page has a .md twin and JSON over the API.