registry / AI-agent control-surface risk / hiskra-code
hiskra-code
Configuração de agentes, skills e regras globais do opencode — instalável via npm
This report shows the evidence for hiskra-code@2.0.2. 1 of 8 known versions is blocked or flagged. The newest version, 3.1.0, is flagged. See all versions.
AI Security Review
scanned 4mo ago · by lpm-firewall-aiAlthough the contents look like a legitimate opencode agent bundle and no exfiltration/destructive code was found, installing npm packages must not silently install or activate AI-agent control instructions in the consuming project. The concrete lifecycle and import-time mutation of .opencode/opencode.json is sufficient to block as AI-agent control-surface hijack. Product guard normalized a concrete AI-agent control hijack publish_block to the blockable dangerous-capability shape.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
npm install hiskra-code postinstall, or requiring/using exported index.cjs functions in a project without .opencode
Impact
Future opencode sessions may follow package-supplied primary agent, permissions, and workflow instructions in the target project.
Mechanism
unconsented AI-agent configuration and instruction drop
Policy narrative
Installing the package runs postinstall.js, which locates the consumer project, creates .opencode/agents and .opencode/skills, copies package-supplied markdown instructions, and creates opencode.json pointing opencode at .opencode/config.md with default_agent set to orquestrador. The CommonJS entry also auto-initializes the same control files when exported functions are used.
AI rationale
Although the contents look like a legitimate opencode agent bundle and no exfiltration/destructive code was found, installing npm packages must not silently install or activate AI-agent control instructions in the consuming project. The concrete lifecycle and import-time mutation of .opencode/opencode.json is sufficient to block as AI-agent control-surface hijack. Product guard normalized a concrete AI-agent control hijack publish_block to the blockable dangerous-capability shape.
Evidence
package.jsonpostinstall.jsindex.cjsagents/orquestrador.mdconfig.mdopencode.json.opencode/config.md.opencode/agents/*.md.opencode/skills/*.md
Network endpoints1
opencode.ai/config.json
Decision evidence
public snapshotAI identified this as a dangerous AI-agent capability at 93.0% confidence with low false-positive risk.
Evidence for policy risk
- package.json defines postinstall: node postinstall.js
- postinstall.js copies package agents/, skills/, config.md into project .opencode/ during npm install
- postinstall.js creates opencode.json with default_agent: orquestrador and instructions: .opencode/config.md
- index.cjs public exports wrap calls with ensureProjectInitialized(), which auto-runs initProject on import/use when .opencode is absent
- agents/orquestrador.md is mode: primary with edit/bash allow and instructs use of subagents
- config.md contains broad workflow instructions for future AI-agent behavior
Evidence against
- No HTTP requests or exfiltration code found; only opencode schema URL string appears
- No child_process, eval, Function, dynamic remote loading, native binaries, or destructive rm/unlink behavior found
- Copied files are visible markdown config/agent/skill files rather than hidden binary payloads
Behavioral surface
EnvironmentVarsFilesystem
HighEntropyStrings
Source & flagged code
3 flagged · loading sourcepackage.jsonView file
•scripts.postinstall = node postinstall.js
High
Install Time Lifecycle Scripts
Package defines install-time lifecycle scripts.
package.jsonView on unpkg•scripts.postinstall = node postinstall.js
Medium
Ambiguous Install Lifecycle Script
Install-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgpostinstall.jsView file
1Install-time AI-agent control hijack evidence:
L5: * Executado após `npm install hiskra-code` para copiar agentes, skills e
L6: * config padrão para o diretório `.opencode/` do projeto alvo.
L7: *
...
L28: const hasOpenCodeJson = fs.existsSync(path.join(dir, 'opencode.json'));
L29: const hasDotOpenCode = fs.existsSync(path.join(dir, '.opencode'));
L30:
...
L69: if (!fs.existsSync(dest)) {
L70: fs.mkdirSync(dest, { recursive: true });
L71: }
...
L83: if (overwrite || !fs.existsSync(destPath)) {
L84: fs.copyFileSync(srcPath, destPath);
L85: count++;
Payload evidence from agents/analista-requisitos.md:
L1: ---
L2: description: Levanta, documenta e valida requisitos funcionais e não funcionais.
Findings
1 Critical1 High3 Medium5 Low
CriticalAi Agent Control Hijackpostinstall.js
HighInstall Time Lifecycle Scriptspackage.json
MediumAmbiguous Install Lifecycle Scriptpackage.json
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
LowNo License
Affected versions and remediation
This report applies to hiskra-code@2.0.2.
The versions list below shows the verdict for each recorded version.
- Avoid installing hiskra-code@2.0.2. Remove it from direct dependencies and check your lockfile for transitive copies.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
- If this version ran, investigate the affected machine and build environment. Rotate credentials it could access and rebuild from a trusted environment.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.