convert-to-cpm— agent skill

convert-to-cpm — independently scanned and version-tracked by SaferSkills.

by dotnet·Agent Skill·github.com/dotnet/skills

Is convert-to-cpm safe to install?

SaferSkills independently audited convert-to-cpm (Agent Skill) and scored it 45/100 (orange). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 7 high-severity and 2 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.

Score
45/100
●●●●●○○○○○
Re-scan~30s
Latest scan
ScannedJun 27, 2026 · 130d ago
Scans run0 over 90 days
Detectors55 checks · 5 categories
Findings2 warnings · 7 high
EngineSaferSkills 2b638c6
View methodology →
SaferSkills installs
This week0
This month0
All time0
CategoryWeightCategory scoreContribution
Securityprompt, exec, net, exfil, eval
35%
0
0.0 pts
Supply chainhash, typosquat, maintainer, lockfile
20%
100
20.0 pts
Maintenancestaleness, pinning, CI
15%
100
15.0 pts
TransparencySKILL.md, perms, README
15%
100
15.0 pts
Communityinstalls, verify, response
15%
100
15.0 pts

Findings & checks · 9 flagged

Securityscore 0 · 9 findings
HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · .github/workflows/devops-health-check.lock.yml
HIGH — once decoded by the agent, an encoded payload has the same impact class as plain-text injection.
Why it matters

A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.

The exact value spotted
excerpt.github/workflows/devops-health-check.lock.yml· yaml
1107setupGlobals(core, github, context, exec, io, getOctokit);
1108const { main } = require(path.join(actionsDir, 'trace_graders.cjs'));
1109await main('eyJ2ZXJzaW9uIjoxLCJncmFkZXJzIjpbeyJpZCI6InRvb2wtc3VjY2Vzcy1yYXRlIiwibmFtZSI6IlRv
… (96 chars elided on L1109)
1110- name: Redact grader outputs
1111if: always()
Occurrences
1 occurrence · at L1109
How to fix
Remove the encoded blob, or decode it and review what it actually contains.
  1. Decode the base64 string and confirm it is not an instruction directed at the agent.
  2. Move any legitimate binary or signature data into a dedicated file (*.sig, SIGNATURES) outside the documentation.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-B64-PAYLOAD-01sha256f834a6cb209b1568…rubric 365aacaView on GitHub
HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · .github/workflows/devops-health-groom.lock.yml
HIGH — once decoded by the agent, an encoded payload has the same impact class as plain-text injection.
Why it matters

A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.

The exact value spotted
excerpt.github/workflows/devops-health-groom.lock.yml· yaml
1096setupGlobals(core, github, context, exec, io, getOctokit);
1097const { main } = require(path.join(actionsDir, 'trace_graders.cjs'));
1098await main('eyJ2ZXJzaW9uIjoxLCJncmFkZXJzIjpbeyJpZCI6InRvb2wtc3VjY2Vzcy1yYXRlIiwibmFtZSI6IlRv
… (96 chars elided on L1098)
1099- name: Redact grader outputs
1100if: always()
Occurrences
1 occurrence · at L1098
How to fix
Remove the encoded blob, or decode it and review what it actually contains.
  1. Decode the base64 string and confirm it is not an instruction directed at the agent.
  2. Move any legitimate binary or signature data into a dedicated file (*.sig, SIGNATURES) outside the documentation.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-B64-PAYLOAD-01sha256f834a6cb209b1568…rubric 365aacaView on GitHub
HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · .github/workflows/devops-health-investigate.lock.yml
HIGH — once decoded by the agent, an encoded payload has the same impact class as plain-text injection.
Why it matters

A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.

The exact value spotted
excerpt.github/workflows/devops-health-investigate.lock.yml· yaml
1147setupGlobals(core, github, context, exec, io, getOctokit);
1148const { main } = require(path.join(actionsDir, 'trace_graders.cjs'));
1149await main('eyJ2ZXJzaW9uIjoxLCJncmFkZXJzIjpbeyJpZCI6InRvb2wtc3VjY2Vzcy1yYXRlIiwibmFtZSI6IlRv
… (96 chars elided on L1149)
1150- name: Redact grader outputs
1151if: always()
Occurrences
1 occurrence · at L1149
How to fix
Remove the encoded blob, or decode it and review what it actually contains.
  1. Decode the base64 string and confirm it is not an instruction directed at the agent.
  2. Move any legitimate binary or signature data into a dedicated file (*.sig, SIGNATURES) outside the documentation.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-B64-PAYLOAD-01sha256f834a6cb209b1568…rubric 365aacaView on GitHub
HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · .github/workflows/issue-investigate.lock.yml
HIGH — once decoded by the agent, an encoded payload has the same impact class as plain-text injection.
Why it matters

A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.

The exact value spotted
excerpt.github/workflows/issue-investigate.lock.yml· yaml
1243setupGlobals(core, github, context, exec, io, getOctokit);
1244const { main } = require(path.join(actionsDir, 'trace_graders.cjs'));
1245await main('eyJ2ZXJzaW9uIjoxLCJncmFkZXJzIjpbeyJpZCI6InRvb2wtc3VjY2Vzcy1yYXRlIiwibmFtZSI6IlRv
… (96 chars elided on L1245)
1246- name: Redact grader outputs
1247if: always()
Occurrences
1 occurrence · at L1245
How to fix
Remove the encoded blob, or decode it and review what it actually contains.
  1. Decode the base64 string and confirm it is not an instruction directed at the agent.
  2. Move any legitimate binary or signature data into a dedicated file (*.sig, SIGNATURES) outside the documentation.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-B64-PAYLOAD-01sha256f834a6cb209b1568…rubric 365aacaView on GitHub
HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · .github/workflows/issue-triage.lock.yml
HIGH — once decoded by the agent, an encoded payload has the same impact class as plain-text injection.
Why it matters

A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.

The exact value spotted
excerpt.github/workflows/issue-triage.lock.yml· yaml
1248setupGlobals(core, github, context, exec, io, getOctokit);
1249const { main } = require(path.join(actionsDir, 'trace_graders.cjs'));
1250await main('eyJ2ZXJzaW9uIjoxLCJncmFkZXJzIjpbeyJpZCI6InRvb2wtc3VjY2Vzcy1yYXRlIiwibmFtZSI6IlRv
… (96 chars elided on L1250)
1251- name: Redact grader outputs
1252if: always()
Occurrences
1 occurrence · at L1250
How to fix
Remove the encoded blob, or decode it and review what it actually contains.
  1. Decode the base64 string and confirm it is not an instruction directed at the agent.
  2. Move any legitimate binary or signature data into a dedicated file (*.sig, SIGNATURES) outside the documentation.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-B64-PAYLOAD-01sha256f834a6cb209b1568…rubric 365aacaView on GitHub
HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · .github/workflows/markdown-linter.lock.yml
HIGH — once decoded by the agent, an encoded payload has the same impact class as plain-text injection.
Why it matters

A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.

The exact value spotted
excerpt.github/workflows/markdown-linter.lock.yml· yaml
1183setupGlobals(core, github, context, exec, io, getOctokit);
1184const { main } = require(path.join(actionsDir, 'trace_graders.cjs'));
1185await main('eyJ2ZXJzaW9uIjoxLCJncmFkZXJzIjpbeyJpZCI6InRvb2wtc3VjY2Vzcy1yYXRlIiwibmFtZSI6IlRv
… (96 chars elided on L1185)
1186- name: Redact grader outputs
1187if: always()
Occurrences
1 occurrence · at L1185
How to fix
Remove the encoded blob, or decode it and review what it actually contains.
  1. Decode the base64 string and confirm it is not an instruction directed at the agent.
  2. Move any legitimate binary or signature data into a dedicated file (*.sig, SIGNATURES) outside the documentation.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-B64-PAYLOAD-01sha256f834a6cb209b1568…rubric 365aacaView on GitHub
HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · .github/workflows/pr-malicious-scan.agent.lock.yml
HIGH — once decoded by the agent, an encoded payload has the same impact class as plain-text injection.
Why it matters

A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.

The exact value spotted
excerpt.github/workflows/pr-malicious-scan.agent.lock.yml· yaml
1217setupGlobals(core, github, context, exec, io, getOctokit);
1218const { main } = require(path.join(actionsDir, 'trace_graders.cjs'));
1219await main('eyJ2ZXJzaW9uIjoxLCJncmFkZXJzIjpbeyJpZCI6InRvb2wtc3VjY2Vzcy1yYXRlIiwibmFtZSI6IlRv
… (96 chars elided on L1219)
1220- name: Redact grader outputs
1221if: always()
Occurrences
1 occurrence · at L1219
How to fix
Remove the encoded blob, or decode it and review what it actually contains.
  1. Decode the base64 string and confirm it is not an instruction directed at the agent.
  2. Move any legitimate binary or signature data into a dedicated file (*.sig, SIGNATURES) outside the documentation.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-B64-PAYLOAD-01sha256f834a6cb209b1568…rubric 365aacaView on GitHub
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · plugins/dotnet-nuget/skills/convert-to-cpm/SKILL.md
MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.
Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted
excerptplugins/dotnet-nuget/skills/convert-to-cpm/SKILL.md· markdown
61### 1. Scope and preflight
62 
63- Resolve the project/solution scope. For a solution, list its projects. For a directory, se
… (108 chars elided on L63)
64- Determine CPM management scopes separately from CLI targets. Group projects that will shar
… (108 chars elided on L64)
65- Check for `packages.config`; if found, switch to Guard mode and stop.
Occurrences
1 occurrence · at L63
How to fix
Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.
  1. Delete blanket "don't ask / no need to confirm" directives from the skill.
  2. If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-DONT-ASK-01sha25614b7852251bf4666…rubric 365aacaView on GitHub
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · plugins/dotnet-nuget/skills/convert-to-cpm/references/audit-complexities.md
MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.
Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted
excerptplugins/dotnet-nuget/skills/convert-to-cpm/references/audit-complexities.md· markdown
32Do not upgrade any package beyond the highest version already in use across the scope — this
… (108 chars elided on L32)
33 
34If the user already supplied a conflict strategy, treat that as the decision and do not ask
… (108 chars elided on L34)
35 
36- **Major version difference**: Emphasize the risk of breaking API changes. Recommend `Versi
… (74 chars elided on L36)
Occurrences
1 occurrence · at L34
How to fix
Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.
  1. Delete blanket "don't ask / no need to confirm" directives from the skill.
  2. If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-DONT-ASK-01sha25612211d7a427c799a…rubric 365aacaView on GitHub
Supply chainscore 100 · 0 findings
✓—All supply chain checks passedNo findings in this category for the latest scan.pass
Maintenancescore 100 · 0 findings
✓—All maintenance checks passedNo findings in this category for the latest scan.pass
Transparencyscore 100 · 0 findings
✓—All transparency checks passedNo findings in this category for the latest scan.pass
Communityscore 100 · 0 findings
✓—All community checks passedNo findings in this category for the latest scan.pass
Vendor response · right of reply
Are you the maintainer? Submit a response →

Audit the pieces. Scan the whole. Decide.

~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.