Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/yeisonrestrepo/code-conductor/cc-implementgit clone --depth 1 https://github.com/yeisonrestrepo/code-conductorWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00011 | $0.02828 |
| Opus 5 | $0.00005 | $0.01414 |
| Sonnet 5 | $0.00002 | $0.00566 |
| Haiku 4.5 | $0.00001 | $0.00283 |
Grade C, and why
cc-implement scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Hidden instructionshighPrompt injection
Directives inside HTML comments, invisible characters or bidirectional overrides are read by the model and not by the person reviewing the file.
3. Strip Unicode invisible characters: ``, ` `, ``, `` **`old_string` construction** (Steps 3 and 5 — the actual string passed to `Edit`): - Take the Read result verbatim; strip trailing whitespace and `\r` only - Pre How it starts
The opening of the file, as written. The whole thing — 165 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Phase entry - Resume Read
Before doing anything else, restore any stored context for the current commit by running scripts/resume-read.mjs. It resolves the current git hash, prefers a valid DB snapshot (conductor-db get-snapshot), falls back to the .claude/memory/session-snapshot.json handoff file, and prints a RESUME_HIT block on a hit / nothing on a miss. Capture its stdout and its exit code with the canonical per-platform form (each first probes for node and treats its absence as a clean miss, never an error):
- Unix / Git Bash:
if command -v node >/dev/null 2>&1; then resume_out="$(node scripts/resume-read.mjs 2>>.conductor/last-write.log)"; resume_rc=$? else resume_rc=3; resume_out=""; fi - PowerShell:
if (Get-Command node -ErrorAction SilentlyContinue) { $__eap = $ErrorActionPreference; $ErrorActionPreference = 'Continue' if (Test-Path variable:PSNativeCommandUseErrorActionPreference) { $__nap = $PSNativeCommandUseErrorActionPreference; $PSNativeCommandUseErrorActionPreference = $false } try { $resume_out = node scripts/resume-read.mjs 2>> .conductor/last-write.log; $resume_rc = $LASTEXITCODE } catch { $resume_rc = 3; $resume_out = "" } finally { $ErrorActionPreference = $__eap if (Test-Path variable:__nap) { $PSNativeCommandUseErrorActionPreference = $__nap } } } else { $resume_rc = 3; $resume_out = "" }
Branch on resume_rc - only 0 and 4 are meaningful; every other code proceeds fresh:
0→ parse the captured block and adopt it as this phase's starting context, then echo one banner to the user:> Resumed from stored snapshot @ <commit> (phase: <phase>), appending(checkpoint prose available)when the block reportsprose: available. Parsing (the command owns normalization): split on\n; strip a trailing\rfrom every line; drop leading/trailing wholly-blank lines; requirelines[0].trim() === 'RESUME_HIT'(anything else = miss);key: valuelines split on the first': '(both sides trimmed); thepending:block is every subsequent^\s*-\s+line up to the first blank line or EOF, each item trimmed. Unknown keys are ignored. In PowerShell,node …bindsstring[]for multi-line output - normalize with$lines = @($resume_out); a$null/empty capture withresume_rc = 3is a miss.4→ operational halt. Do not run this phase's normal work. Emit exactly:SNAP_INVALID: corrupt handoff at .claude/memory/session-snapshot.json - inspect or remove it, then re-run.and enter standby awaiting user action. The corrupt file is left on disk (the script did not delete it).3or any other code → proceed fresh (clean miss, bypass, degrade, absentnode, or any unexpected runtime code). Ignore the capture.
resume-read.mjs writes its own trace lines to .conductor/last-write.log via appendFileSync; the 2>> redirect above only sinks the incidental exit-4 halt reason away from the UI - it is not the trace channel.
Read .claude/memory/project.md before starting any task.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 165 lines · 11 tokens per session scan C 1c56b4980a3e
cc-implement is a command published in the GitHub repository yeisonrestrepo/code-conductor (6 stars, last pushed 1mo ago), licensed MIT. It adds 11 tokens to every session and 2,828 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it C with 1 finding (hidden instructions). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
edit
Edit a stash item — inline replacement, or open in editor (preserves timestamp).
pop
Pop an item off the stash and continue working on it in the current session.
drop
Drop a stash item without working on it.
log
Show history of popped and dropped stash items.
show
Show a single stash item by number (for reading or copying its text).
list
Show all items currently in the stash.