Back to skill

Security audit

Funding Program Manager

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent Karma funding-program manager, but it handles high-impact admin and payout actions with persistent plaintext credentials and an unrestricted API host override.

Review this skill carefully before installing. Only use it with a Karma API key you are willing to grant to the agent, avoid setting KARMA_API_URL unless you fully trust the endpoint, prefer session-only credentials or secure secret storage, and manually confirm approvals, payouts, reviewer changes, agreement signatures, and AI evaluation submissions before they are sent.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:18
Finding

API Key and Sensitive Data Exfiltration Through an Unrestricted API Host Override

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:66
Finding

Karma API Key Stored in Plaintext Without Restrictive File Permissions

Content
View full analysis
"${CLAUDE_PLUGIN_DATA}/credentials.json" export KARMA_API_KEY="karma_..." ``` ```bash if [ -f "$HOME/.zshrc" ]; then SHELL_RC="$HOME/.zshrc" elif [ -f "$HOME/.bashrc" ]; then SHELL_RC="$HOME/.bashrc"; fi grep -q 'KARMA_API_KEY' "$SHELL_RC" 2>/dev/null && sed -i.bak 's/export KARMA_API_KEY=.*/export KARMA_API_KEY="karma_..."/' "$SHELL_RC" || echo '\n# Karma API Key\nexport KARMA_API_KEY="karma_..."' >> "$SHELL_RC" export KARMA_API_KEY="karma_..." ``` ### Technical Analysis The setup instructions persist a reusable API key directly into either `credentials.json` or a shell startup file. They do not set a restrictive `umask`, verify ownership, or apply mode `0600` to the resulting file. File permissions therefore depend on the user's environment and existing directory settings. The shell configuration path is also broader than necessary because every future interactive shell loads the secret into its environment. Such environment variables may be inherited by unrelated child processes or exposed through diagnostics and process-inspection mechanisms available under the local security model. The `sed -i.bak` command creates a backup of the previous shell configuration. If a key was already present, the backup may retain the old credential even after replacement or rotation. This increases the number and lifetime of plaintext secret copies. ### Attack Path 1. A user follows the Skill's automatic API-key setup flow. 2. The Skill writes the API key to `credentials.json` or a shell startup file using the environment's default permissions. 3. In the shell-update branch, `sed -i.bak` may create another plaintext copy in a backup file. 4. Another local user, compromised process, backup collector, support bundle, ...[truncated 783 chars]
Remediation
View remediation
"${CLAUDE_PLUGIN_DATA}/credentials.json" chmod 600 "${CLAUDE_PLUGIN_DATA}/credentials.json" ``` 3. Verify that the credential directory and file are owned by the current user and reject symlinks before writing. 4. Do not persist the key to `.zshrc` or `.bashrc`; load it only for the process that requires it. 5. Avoid `sed -i.bak` for secret-bearing files. Securely remove any existing backup after migration. 6. Prevent the credential file from being included in source control, synchronization, telemetry, logs, or diagnostic bundles. 7. Provide explicit key revocation and rotation instructions. 8. Prefer short-lived, narrowly scoped tokens over long-lived API keys, especially for payout and administrative operations. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (12)

Credential Access

High
Category
Privilege Escalation
Confidence
98% confidence
Finding

The skill directs writing an API key into a local credentials.json file, which is a credential materialization and persistence action with significant exposure risk. If the file is readable by other processes, included in backups, synced, or mishandled by surrounding tooling, the key could be stolen and used to administer programs, view application data, and perform privileged operations.

Content

Scanner excerpt · SKILL.md (reported line 71)May include surrounding context.

bash
mkdir -p "${CLAUDE_PLUGIN_DATA}"
echo '{"apiKey": "karma_..."}' > "${CLAUDE_PLUGIN_DATA}/credentials.json"
export KARMA_API_KEY="karma_..."

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The trigger phrase "or any funding program administration action" is overly broad and can cause the skill to activate for a wide range of ambiguous requests. In a high-privilege admin skill that can create programs, approve applications, and disburse funds, over-triggering increases the chance of unintended sensitive actions being proposed or initiated in the wrong context.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 46)May include surrounding context.

Quick Start (No Account Needed)

bash
curl -s -X POST "${BASE_URL}/v2/agent/register" \
  -H "Content-Type: application/json" \
  -H "X-Source: skill:funding-program-manager" -H "X-Invocation-Id: $INVOCATION_ID" -H "X-Skill-Version: 1.0.0" \
  -d '{}'

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill instructs automatic persistence of an API key into plugin storage or shell startup files, which creates long-lived credentials on disk and modifies user configuration. This is dangerous because it increases the exposure window for credential theft, can leak to other local processes or users, and performs persistent system changes without a strong explicit consent/warning step.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 87)May include surrounding context.

Verify

bash
curl -s "${BASE_URL}/v2/agent/info" \
  -H "x-api-key: ${KARMA_API_KEY}" \
  -H "X-Source: skill:funding-program-manager" -H "X-Invocation-Id: $INVOCATION_ID" -H "X-Skill-Version: 1.0.0"

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 177)May include surrounding context.

After the program exists in the registry, create its funding configuration to enable application management.

bash
curl -s -X POST "${BASE_URL}/v2/funding-program-configs/${PROGRAM_ID}" \
  -H "Content-Type: application/json" \
  -H "x-api-key: ${API_KEY}" \
  -H "X-Source: skill:funding-program-manager" -H "X-Invocation-Id: $INVOCATION_ID" -H "X-Skill-Version: 1.0.0" \

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 202)May include surrounding context.

Define the fields applicants must fill out. The form must contain at least one email field.

bash
curl -s -X PUT "${BASE_URL}/v2/funding-program-configs/${PROGRAM_ID}" \
  -H "Content-Type: application/json" \
  -H "x-api-key: ${API_KEY}" \
  -H "X-Source: skill:funding-program-manager" -H "X-Invocation-Id: $INVOCATION_ID" -H "X-Skill-Version: 1.0.0" \

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 285)May include surrounding context.

Get Program Details

bash
curl -s "${BASE_URL}/v2/funding-program-configs/${PROGRAM_ID}" \
  -H "x-api-key: ${API_KEY}" \
  -H "X-Source: skill:funding-program-manager" -H "X-Invocation-Id: $INVOCATION_ID" -H "X-Skill-Version: 1.0.0"

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 369)May include surrounding context.

Get Application Details

bash
curl -s "${BASE_URL}/v2/funding-applications/${REFERENCE_NUMBER}" \
  -H "x-api-key: ${API_KEY}" \
  -H "X-Source: skill:funding-program-manager" -H "X-Invocation-Id: $INVOCATION_ID" -H "X-Skill-Version: 1.0.0"

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The application workflow collects applicant email and detailed application contents, then sends them to remote APIs and potentially to AI evaluation endpoints, but does not require a clear user-facing disclosure before transmission. This creates privacy and consent risk because users may reveal personal or sensitive project information without understanding that it will be processed remotely and reused for evaluation.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
92% confidence
Finding

This endpoint transmits user-provided application answers to a remote AI evaluation service, potentially including personal and sensitive project information, without a mandatory explicit consent step in the flow. Because the skill invites users to submit free-form content for scoring, the privacy impact is materially higher than a generic API call and can expose confidential information to automated processing.

Content

Scanner excerpt · SKILL.md (reported line 445)May include surrounding context.

If enabled, call the evaluate-realtime endpoint with the user's answers:

bash
curl -s -X POST "${BASE_URL}/v2/funding-applications/${PROGRAM_ID}/evaluate-realtime" \
  -H "Content-Type: application/json" \
  -H "x-api-key: ${API_KEY}" \
  -H "X-Source: skill:funding-program-manager" -H "X-Invocation-Id: $INVOCATION_ID" -H "X-Skill-Version: 1.0.0" \

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 558)May include surrounding context.

List Program Reviewers

bash
curl -s "${BASE_URL}/v2/funding-program-configs/${PROGRAM_ID}/reviewers" \
  -H "x-api-key: ${API_KEY}" \
  -H "X-Source: skill:funding-program-manager" -H "X-Invocation-Id: $INVOCATION_ID" -H "X-Skill-Version: 1.0.0"